修复retry会清空次数的问题、优化流程阈值，更新版本号beta06.03

Feature/new algorithm

.gitignore

@@ -34,6 +34,394 @@
 Debug
 Release
 .vs
-*.vcxproj.filters
 *.vcxproj.user
 *.swp
+
+## Ignore Visual Studio temporary files, build results, and
+## files generated by popular Visual Studio add-ons.
+##
+## Get latest from https://github.com/github/gitignore/blob/master/VisualStudio.gitignore
+
+# User-specific files
+*.rsuser
+*.suo
+*.user
+*.userosscache
+*.sln.docstates
+
+# User-specific files (MonoDevelop/Xamarin Studio)
+*.userprefs
+
+# Mono auto generated files
+mono_crash.*
+
+# Build results
+[Dd]ebug/
+[Dd]ebugPublic/
+[Rr]elease/
+[Rr]eleases/
+x64/
+x86/
+[Ww][Ii][Nn]32/
+[Aa][Rr][Mm]/
+[Aa][Rr][Mm]64/
+bld/
+[Bb]in/
+[Oo]bj/
+[Ll]og/
+[Ll]ogs/
+
+# Visual Studio 2015/2017 cache/options directory
+.vs/
+# Uncomment if you have tasks that create the project's static files in wwwroot
+#wwwroot/
+
+# Visual Studio 2017 auto generated files
+Generated\ Files/
+
+# MSTest test Results
+[Tt]est[Rr]esult*/
+[Bb]uild[Ll]og.*
+
+# NUnit
+*.VisualState.xml
+TestResult.xml
+nunit-*.xml
+
+# Build Results of an ATL Project
+[Dd]ebugPS/
+[Rr]eleasePS/
+dlldata.c
+
+# Benchmark Results
+BenchmarkDotNet.Artifacts/
+
+# .NET Core
+project.lock.json
+project.fragment.lock.json
+artifacts/
+
+# ASP.NET Scaffolding
+ScaffoldingReadMe.txt
+
+# StyleCop
+StyleCopReport.xml
+
+# Files built by Visual Studio
+*_i.c
+*_p.c
+*_h.h
+*.ilk
+*.meta
+*.obj
+*.iobj
+*.pch
+*.pdb
+*.ipdb
+*.pgc
+*.pgd
+*.rsp
+*.sbr
+*.tlb
+*.tli
+*.tlh
+*.tmp
+*.tmp_proj
+*_wpftmp.csproj
+*.log
+*.tlog
+*.vspscc
+*.vssscc
+.builds
+*.pidb
+*.svclog
+*.scc
+
+# Chutzpah Test files
+_Chutzpah*
+
+# Visual C++ cache files
+ipch/
+*.aps
+*.ncb
+*.opendb
+*.opensdf
+*.sdf
+*.cachefile
+*.VC.db
+*.VC.VC.opendb
+
+# Visual Studio profiler
+*.psess
+*.vsp
+*.vspx
+*.sap
+
+# Visual Studio Trace Files
+*.e2e
+
+# TFS 2012 Local Workspace
+$tf/
+
+# Guidance Automation Toolkit
+*.gpState
+
+# ReSharper is a .NET coding add-in
+_ReSharper*/
+*.[Rr]e[Ss]harper
+*.DotSettings.user
+
+# TeamCity is a build add-in
+_TeamCity*
+
+# DotCover is a Code Coverage Tool
+*.dotCover
+
+# AxoCover is a Code Coverage Tool
+.axoCover/*
+!.axoCover/settings.json
+
+# Coverlet is a free, cross platform Code Coverage Tool
+coverage*.json
+coverage*.xml
+coverage*.info
+
+# Visual Studio code coverage results
+*.coverage
+*.coveragexml
+
+# NCrunch
+_NCrunch_*
+.*crunch*.local.xml
+nCrunchTemp_*
+
+# MightyMoose
+*.mm.*
+AutoTest.Net/
+
+# Web workbench (sass)
+.sass-cache/
+
+# Installshield output folder
+[Ee]xpress/
+
+# DocProject is a documentation generator add-in
+DocProject/buildhelp/
+DocProject/Help/*.HxT
+DocProject/Help/*.HxC
+DocProject/Help/*.hhc
+DocProject/Help/*.hhk
+DocProject/Help/*.hhp
+DocProject/Help/Html2
+DocProject/Help/html
+
+# Click-Once directory
+publish/
+
+# Publish Web Output
+*.[Pp]ublish.xml
+*.azurePubxml
+# Note: Comment the next line if you want to checkin your web deploy settings,
+# but database connection strings (with potential passwords) will be unencrypted
+*.pubxml
+*.publishproj
+
+# Microsoft Azure Web App publish settings. Comment the next line if you want to
+# checkin your Azure Web App publish settings, but sensitive information contained
+# in these scripts will be unencrypted
+PublishScripts/
+
+# NuGet Packages
+*.nupkg
+# NuGet Symbol Packages
+*.snupkg
+# The packages folder can be ignored because of Package Restore
+**/[Pp]ackages/*
+# except build/, which is used as an MSBuild target.
+!**/[Pp]ackages/build/
+# Uncomment if necessary however generally it will be regenerated when needed
+#!**/[Pp]ackages/repositories.config
+# NuGet v3's project.json files produces more ignorable files
+*.nuget.props
+*.nuget.targets
+
+# Nuget personal access tokens and Credentials
+nuget.config
+
+# Microsoft Azure Build Output
+csx/
+*.build.csdef
+
+# Microsoft Azure Emulator
+ecf/
+rcf/
+
+# Windows Store app package directories and files
+AppPackages/
+BundleArtifacts/
+Package.StoreAssociation.xml
+_pkginfo.txt
+*.appx
+*.appxbundle
+*.appxupload
+
+# Visual Studio cache files
+# files ending in .cache can be ignored
+*.[Cc]ache
+# but keep track of directories ending in .cache
+!?*.[Cc]ache/
+
+# Others
+ClientBin/
+~$*
+*~
+*.dbmdl
+*.dbproj.schemaview
+*.jfm
+*.pfx
+*.publishsettings
+orleans.codegen.cs
+
+# Including strong name files can present a security risk
+# (https://github.com/github/gitignore/pull/2483#issue-259490424)
+#*.snk
+
+# Since there are multiple workflows, uncomment next line to ignore bower_components
+# (https://github.com/github/gitignore/pull/1529#issuecomment-104372622)
+#bower_components/
+
+# RIA/Silverlight projects
+Generated_Code/
+
+# Backup & report files from converting an old project file
+# to a newer Visual Studio version. Backup files are not needed,
+# because we have git ;-)
+_UpgradeReport_Files/
+Backup*/
+UpgradeLog*.XML
+UpgradeLog*.htm
+ServiceFabricBackup/
+*.rptproj.bak
+
+# SQL Server files
+*.mdf
+*.ldf
+*.ndf
+
+# Business Intelligence projects
+*.rdl.data
+*.bim.layout
+*.bim_*.settings
+*.rptproj.rsuser
+*- [Bb]ackup.rdl
+*- [Bb]ackup ([0-9]).rdl
+*- [Bb]ackup ([0-9][0-9]).rdl
+
+# Microsoft Fakes
+FakesAssemblies/
+
+# GhostDoc plugin setting file
+*.GhostDoc.xml
+
+# Node.js Tools for Visual Studio
+.ntvs_analysis.dat
+node_modules/
+
+# Visual Studio 6 build log
+*.plg
+
+# Visual Studio 6 workspace options file
+*.opt
+
+# Visual Studio 6 auto-generated workspace file (contains which files were open etc.)
+*.vbw
+
+# Visual Studio LightSwitch build output
+**/*.HTMLClient/GeneratedArtifacts
+**/*.DesktopClient/GeneratedArtifacts
+**/*.DesktopClient/ModelManifest.xml
+**/*.Server/GeneratedArtifacts
+**/*.Server/ModelManifest.xml
+_Pvt_Extensions
+
+# Paket dependency manager
+.paket/paket.exe
+paket-files/
+
+# FAKE - F# Make
+.fake/
+
+# CodeRush personal settings
+.cr/personal
+
+# Python Tools for Visual Studio (PTVS)
+__pycache__/
+*.pyc
+
+# Cake - Uncomment if you are using it
+# tools/**
+# !tools/packages.config
+
+# Tabs Studio
+*.tss
+
+# Telerik's JustMock configuration file
+*.jmconfig
+
+# BizTalk build output
+*.btp.cs
+*.btm.cs
+*.odx.cs
+*.xsd.cs
+
+# OpenCover UI analysis results
+OpenCover/
+
+# Azure Stream Analytics local run output
+ASALocalRun/
+
+# MSBuild Binary and Structured Log
+*.binlog
+
+# NVidia Nsight GPU debugger configuration file
+*.nvuser
+
+# MFractors (Xamarin productivity tool) working folder
+.mfractor/
+
+# Local History for Visual Studio
+.localhistory/
+
+# BeatPulse healthcheck temp database
+healthchecksdb
+
+# Backup folder for Package Reference Convert tool in Visual Studio 2017
+MigrationBackup/
+
+# Ionide (cross platform F# VS Code tools) working folder
+.ionide/
+
+# Fody - auto-generated XML schema
+FodyWeavers.xsd
+
+# VS Code files for those working on multiple tools
+.vscode/*
+!.vscode/settings.json
+!.vscode/tasks.json
+!.vscode/launch.json
+!.vscode/extensions.json
+*.code-workspace
+
+# Local History for Visual Studio Code
+.history/
+
+# Windows Installer files from build outputs
+*.cab
+*.msi
+*.msix
+*.msm
+*.msp
+
+# JetBrains Rider
+.idea/
+*.sln.iml

.gitmodules

@@ -1,3 +0,0 @@
-[submodule "meojson"]
-	path = meojson
-	url = https://github.com/MistEO/meojson.git

3rdPart/include/OcrLiteOnnx/AngleNet.h

@@ -0,0 +1,41 @@
+#ifndef __OCR_ANGLENET_H__
+#define __OCR_ANGLENET_H__
+
+#include "OcrStruct.h"
+#include "onnxruntime_cxx_api.h"
+#include <opencv/cv.hpp>
+
+class AngleNet {
+public:
+    AngleNet();
+
+    ~AngleNet();
+
+    void setNumThread(int numOfThread);
+
+    void initModel(const std::string &pathStr);
+
+    std::vector<Angle> getAngles(std::vector<cv::Mat> &partImgs, const char *path,
+                                 const char *imgName, bool doAngle, bool mostAngle);
+
+private:
+    bool isOutputAngleImg = false;
+
+    Ort::Session *session;
+    Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_ERROR, "AngleNet");
+    Ort::SessionOptions sessionOptions = Ort::SessionOptions();
+    int numThread = 0;
+
+    char *inputName;
+    char *outputName;
+
+    const float meanValues[3] = {127.5, 127.5, 127.5};
+    const float normValues[3] = {1.0 / 127.5, 1.0 / 127.5, 1.0 / 127.5};
+    const int dstWidth = 192;
+    const int dstHeight = 32;
+
+    Angle getAngle(cv::Mat &src);
+};
+
+
+#endif //__OCR_ANGLENET_H__

3rdPart/include/OcrLiteOnnx/CrnnNet.h

@@ -0,0 +1,43 @@
+#ifndef __OCR_CRNNNET_H__
+#define __OCR_CRNNNET_H__
+
+#include "OcrStruct.h"
+#include "onnxruntime_cxx_api.h"
+#include <opencv/cv.hpp>
+
+class CrnnNet {
+public:
+
+    CrnnNet();
+
+    ~CrnnNet();
+
+    void setNumThread(int numOfThread);
+
+    void initModel(const std::string &pathStr, const std::string &keysPath);
+
+    std::vector<TextLine> getTextLines(std::vector<cv::Mat> &partImg, const char *path, const char *imgName);
+
+private:
+    bool isOutputDebugImg = false;
+    Ort::Session *session;
+    Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_ERROR, "CrnnNet");
+    Ort::SessionOptions sessionOptions = Ort::SessionOptions();
+    int numThread = 0;
+
+    char *inputName;
+    char *outputName;
+
+    const float meanValues[3] = {127.5, 127.5, 127.5};
+    const float normValues[3] = {1.0 / 127.5, 1.0 / 127.5, 1.0 / 127.5};
+    const int dstHeight = 32;
+
+    std::vector<std::string> keys;
+
+    TextLine scoreToTextLine(const std::vector<float> &outputData, int h, int w);
+
+    TextLine getTextLine(const cv::Mat &src);
+};
+
+
+#endif //__OCR_CRNNNET_H__

3rdPart/include/OcrLiteOnnx/DbNet.h

@@ -0,0 +1,34 @@
+#ifndef __OCR_DBNET_H__
+#define __OCR_DBNET_H__
+
+#include "OcrStruct.h"
+#include "onnxruntime_cxx_api.h"
+#include <opencv/cv.hpp>
+
+class DbNet {
+public:
+    DbNet();
+
+    ~DbNet();
+
+    void setNumThread(int numOfThread);
+
+    void initModel(const std::string &pathStr);
+
+    std::vector<TextBox> getTextBoxes(cv::Mat &src, ScaleParam &s, float boxScoreThresh,
+                                      float boxThresh, float unClipRatio);
+
+private:
+    Ort::Session *session;
+    Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_ERROR, "DbNet");
+    Ort::SessionOptions sessionOptions = Ort::SessionOptions();
+    int numThread = 0;
+    char *inputName;
+    char *outputName;
+
+    const float meanValues[3] = {0.485 * 255, 0.456 * 255, 0.406 * 255};
+    const float normValues[3] = {1.0 / 0.229 / 255.0, 1.0 / 0.224 / 255.0, 1.0 / 0.225 / 255.0};
+};
+
+
+#endif //__OCR_DBNET_H__

3rdPart/include/OcrLiteOnnx/OcrLite.h

@@ -0,0 +1,56 @@
+#ifndef __OCR_LITE_H__
+#define __OCR_LITE_H__
+
+#include "opencv2/core.hpp"
+#include "onnxruntime_cxx_api.h"
+#include "OcrStruct.h"
+#include "DbNet.h"
+#include "AngleNet.h"
+#include "CrnnNet.h"
+
+class OcrLite {
+public:
+    OcrLite();
+
+    ~OcrLite();
+
+    void setNumThread(int numOfThread);
+
+    void initLogger(bool isConsole, bool isPartImg, bool isResultImg);
+
+    void enableResultTxt(const char *path, const char *imgName);
+
+    void initModels(const std::string &detPath, const std::string &clsPath,
+                    const std::string &recPath, const std::string &keysPath);
+
+    void Logger(const char *format, ...);
+
+    OcrResult detect(const char *path, const char *imgName,
+                     int padding, int maxSideLen,
+                     float boxScoreThresh, float boxThresh, float unClipRatio, bool doAngle, bool mostAngle);
+
+
+	OcrResult detect(const cv::Mat& mat,
+		int padding, int maxSideLen,
+		float boxScoreThresh, float boxThresh, float unClipRatio, bool doAngle, bool mostAngle);
+
+private:
+    bool isOutputConsole = false;
+    bool isOutputPartImg = false;
+    bool isOutputResultTxt = false;
+    bool isOutputResultImg = false;
+    FILE *resultTxt;
+    DbNet dbNet;
+    AngleNet angleNet;
+    CrnnNet crnnNet;
+
+    std::vector<cv::Mat> getPartImages(cv::Mat &src, std::vector<TextBox> &textBoxes,
+                                       const char *path, const char *imgName);
+
+    OcrResult detect(const char *path, const char *imgName,
+                     cv::Mat &src, cv::Rect &originRect, ScaleParam &scale,
+                     float boxScoreThresh = 0.6f, float boxThresh = 0.3f,
+                     float unClipRatio = 2.0f, bool doAngle = true, bool mostAngle = true);
+};
+
+#endif //__OCR_LITE_H__

3rdPart/include/OcrLiteOnnx/OcrLiteCaller.h

@@ -0,0 +1,35 @@
+#pragma once
+
+#include <memory>
+#include <string>
+
+#include "OcrLitePort.h"
+#include "OcrStruct.h"
+
+namespace cv
+{
+	class Mat;
+}
+class OcrLite;
+
+class OCRLITE_PORT OcrLiteCaller
+{
+public:
+	OcrLiteCaller();
+	~OcrLiteCaller() = default;
+	OcrLiteCaller(const OcrLite&) = delete;
+	OcrLiteCaller(OcrLite&&) = delete;
+
+	void setNumThread(int numOfThread);
+	void initModels(const std::string& detPath, const std::string& clsPath,
+		const std::string& recPath, const std::string& keysPath);
+
+	OcrResult detect(const cv::Mat& mat,
+		int padding, int maxSideLen,
+		float boxScoreThresh, float boxThresh, float unClipRatio, bool doAngle, bool mostAngle);
+
+	OcrLiteCaller& operator=(const OcrLiteCaller&) = delete;
+	OcrLiteCaller& operator=(OcrLiteCaller&&) = delete;
+private:
+	std::shared_ptr<OcrLite> m_ocrlite_ptr;
+};

3rdPart/include/OcrLiteOnnx/OcrLitePort.h

@@ -0,0 +1,37 @@
+#pragma once
+
+#pragma once
+
+// The way how the function is called
+#if !defined(OCRLITE_CALL)
+#if defined(_WIN32)
+#define OCRLITE_CALL __stdcall
+#else
+#define OCRLITE_CALL
+#endif /* _WIN32 */
+#endif /* ISSCALL */
+
+#if defined _WIN32 || defined __CYGWIN__
+#define OCRLITE_EXPORT __declspec(dllexport)
+#define OCRLITE_IMPORT __declspec(dllimport)
+#define OCRLITE_LOCAL
+#else   // ! defined _WIN32 || defined __CYGWIN__
+#if __GNUC__ >= 4
+#define OCRLITE_EXPORT __attribute__ ((visibility ("default")))
+#define OCRLITE_IMPORT __attribute__ ((visibility ("default")))
+#define OCRLITE_LOCAL  __attribute__ ((visibility ("hidden")))
+#else   // ! __GNUC__ >= 4
+#define OCRLITE_EXPORT
+#define OCRLITE_IMPORT
+#endif  // End __GNUC__ >= 4
+#endif  // End defined _WIN32 || defined __CYGWIN__
+
+#ifdef __CLIB__
+#define OCRLITE_PORT OCRLITE_EXPORT
+#else
+#define OCRLITE_PORT OCRLITE_IMPORT
+#endif // OCRLITE_PORT
+
+#define OCR_API OCRLITE_PORT OCRLITE_CALL
+
+#define OCR_LOCAL OCRLITE_LOCAL OCRLITE_CALL

3rdPart/include/OcrLiteOnnx/OcrResultUtils.h

@@ -0,0 +1,35 @@
+#ifdef __JNI__
+#ifndef __OCR_RESULT_UTILS_H__
+#define __OCR_RESULT_UTILS_H__
+#include <jni.h>
+#include "OcrStruct.h"
+
+class OcrResultUtils {
+public:
+    OcrResultUtils(JNIEnv *env, OcrResult &ocrResult);
+
+    ~OcrResultUtils();
+
+    jobject getJObject();
+
+private:
+    JNIEnv *jniEnv;
+    jobject jOcrResult;
+
+    jclass newJListClass();
+
+    jmethodID getListConstructor(jclass clazz);
+
+    jobject getTextBlock(TextBlock &textBlock);
+
+    jobject getTextBlocks(std::vector<TextBlock> &textBlocks);
+
+    jobject newJPoint(cv::Point &point);
+
+    jobject newJBoxPoint(std::vector<cv::Point> &boxPoint);
+
+    jfloatArray newJScoreArray(std::vector<float> &scores);
+
+};
+#endif //__OCR_RESULT_UTILS_H__
+#endif

3rdPart/include/OcrLiteOnnx/OcrStruct.h

@@ -0,0 +1,55 @@
+#ifndef __OCR_STRUCT_H__
+#define __OCR_STRUCT_H__
+
+#include "opencv2/core.hpp"
+#include <vector>
+
+#include "OcrLitePort.h"
+
+struct ScaleParam {
+    int srcWidth;
+    int srcHeight;
+    int dstWidth;
+    int dstHeight;
+    float ratioWidth;
+    float ratioHeight;
+};
+
+struct TextBox {
+    std::vector<cv::Point> boxPoint;
+    float score;
+};
+
+struct Angle {
+    int index;
+    float score;
+    double time;
+};
+
+struct TextLine {
+    std::string text;
+    std::vector<float> charScores;
+    double time;
+};
+
+struct OCRLITE_PORT TextBlock {
+    std::vector<cv::Point> boxPoint;
+    float boxScore;
+    int angleIndex;
+    float angleScore;
+    double angleTime;
+    std::string text;
+    std::vector<float> charScores;
+    double crnnTime;
+    double blockTime;
+};
+
+struct OCRLITE_PORT OcrResult {
+    double dbNetTime;
+    std::vector<TextBlock> textBlocks;
+    cv::Mat boxImg;
+    double detectTime;
+    std::string strRes;
+};
+
+#endif //__OCR_STRUCT_H__

3rdPart/include/OcrLiteOnnx/OcrUtils.h

@@ -0,0 +1,103 @@
+#ifndef __OCR_UTILS_H__
+#define __OCR_UTILS_H__
+
+#include <opencv2/core.hpp>
+#include "OcrStruct.h"
+#include "onnxruntime_cxx_api.h"
+#include <numeric>
+#include <sys/stat.h>
+
+template<typename T, typename... Ts>
+static std::unique_ptr<T> makeUnique(Ts &&... params) {
+    return std::unique_ptr<T>(new T(std::forward<Ts>(params)...));
+}
+
+template<typename T>
+static double getMean(std::vector<T> &input) {
+    auto sum = accumulate(input.begin(), input.end(), 0.0);
+    return sum / input.size();
+}
+
+template<typename T>
+static double getStdev(std::vector<T> &input, double mean) {
+    if (input.size() <= 1) return 0;
+    double accum = 0.0;
+    for_each(input.begin(), input.end(), [&](const double d) {
+        accum += (d - mean) * (d - mean);
+    });
+    double stdev = sqrt(accum / (input.size() - 1));
+    return stdev;
+}
+
+double getCurrentTime();
+
+inline bool isFileExists(const std::string &name) {
+    struct stat buffer;
+    return (stat(name.c_str(), &buffer) == 0);
+}
+
+#ifdef _WIN32
+#define my_strtol wcstol
+#define my_strrchr wcsrchr
+#define my_strcasecmp _wcsicmp
+#define my_strdup _strdup
+#else
+#define my_strtol strtol
+#define my_strrchr strrchr
+#define my_strcasecmp strcasecmp
+#define my_strdup strdup
+#endif
+
+std::wstring strToWstr(std::string str);
+
+ScaleParam getScaleParam(cv::Mat &src, const float scale);
+
+ScaleParam getScaleParam(cv::Mat &src, const int targetSize);
+
+std::vector<cv::Point2f> getBox(const cv::RotatedRect &rect);
+
+int getThickness(cv::Mat &boxImg);
+
+void drawTextBox(cv::Mat &boxImg, cv::RotatedRect &rect, int thickness);
+
+void drawTextBox(cv::Mat &boxImg, const std::vector<cv::Point> &box, int thickness);
+
+void drawTextBoxes(cv::Mat &boxImg, std::vector<TextBox> &textBoxes, int thickness);
+
+cv::Mat matRotateClockWise180(cv::Mat src);
+
+cv::Mat matRotateClockWise90(cv::Mat src);
+
+cv::Mat getRotateCropImage(const cv::Mat &src, std::vector<cv::Point> box);
+
+cv::Mat adjustTargetImg(cv::Mat &src, int dstWidth, int dstHeight);
+
+std::vector<cv::Point> getMinBoxes(const std::vector<cv::Point> &inVec, float &minSideLen, float &allEdgeSize);
+
+float boxScoreFast(const cv::Mat &inMat, const std::vector<cv::Point> &inBox);
+
+std::vector<cv::Point> unClip(const std::vector<cv::Point> &inBox, float perimeter, float unClipRatio);
+
+std::vector<float> substractMeanNormalize(cv::Mat &src, const float *meanVals, const float *normVals);
+
+std::vector<int> getAngleIndexes(std::vector<Angle> &angles);
+
+std::vector<char *> getInputNames(Ort::Session *session);
+
+std::vector<char *> getOutputNames(Ort::Session *session);
+
+void getInputName(Ort::Session *session, char *&inputName);
+
+void getOutputName(Ort::Session *session, char *&outputName);
+
+void saveImg(cv::Mat &img, const char *imgPath);
+
+std::string getSrcImgFilePath(const char *path, const char *imgName);
+
+std::string getResultTxtFilePath(const char *path, const char *imgName);
+
+std::string getResultImgFilePath(const char *path, const char *imgName);
+
+std::string getDebugImgFilePath(const char *path, const char *imgName, int i, const char *tag);
+
+#endif //__OCR_UTILS_H__

3rdPart/include/OcrLiteOnnx/clipper.hpp

@@ -0,0 +1,406 @@
+/*******************************************************************************
+*                                                                              *
+* Author    :  Angus Johnson                                                   *
+* Version   :  6.4.2                                                           *
+* Date      :  27 February 2017                                                *
+* Website   :  http://www.angusj.com                                           *
+* Copyright :  Angus Johnson 2010-2017                                         *
+*                                                                              *
+* License:                                                                     *
+* Use, modification & distribution is subject to Boost Software License Ver 1. *
+* http://www.boost.org/LICENSE_1_0.txt                                         *
+*                                                                              *
+* Attributions:                                                                *
+* The code in this library is an extension of Bala Vatti's clipping algorithm: *
+* "A generic solution to polygon clipping"                                     *
+* Communications of the ACM, Vol 35, Issue 7 (July 1992) pp 56-63.             *
+* http://portal.acm.org/citation.cfm?id=129906                                 *
+*                                                                              *
+* Computer graphics and geometric modeling: implementation and algorithms      *
+* By Max K. Agoston                                                            *
+* Springer; 1 edition (January 4, 2005)                                        *
+* http://books.google.com/books?q=vatti+clipping+agoston                       *
+*                                                                              *
+* See also:                                                                    *
+* "Polygon Offsetting by Computing Winding Numbers"                            *
+* Paper no. DETC2005-85513 pp. 565-575                                         *
+* ASME 2005 International Design Engineering Technical Conferences             *
+* and Computers and Information in Engineering Conference (IDETC/CIE2005)      *
+* September 24-28, 2005 , Long Beach, California, USA                          *
+* http://www.me.berkeley.edu/~mcmains/pubs/DAC05OffsetPolygon.pdf              *
+*                                                                              *
+*******************************************************************************/
+
+#ifndef clipper_hpp
+#define clipper_hpp
+
+#define CLIPPER_VERSION "6.4.2"
+
+//use_int32: When enabled 32bit ints are used instead of 64bit ints. This
+//improve performance but coordinate values are limited to the range +/- 46340
+//#define use_int32
+
+//use_xyz: adds a Z member to IntPoint. Adds a minor cost to perfomance.
+//#define use_xyz
+
+//use_lines: Enables line clipping. Adds a very minor cost to performance.
+#define use_lines
+
+//use_deprecated: Enables temporary support for the obsolete functions
+//#define use_deprecated  
+
+#include <vector>
+#include <list>
+#include <set>
+#include <stdexcept>
+#include <cstring>
+#include <cstdlib>
+#include <ostream>
+#include <functional>
+#include <queue>
+
+namespace ClipperLib {
+
+    enum ClipType { ctIntersection, ctUnion, ctDifference, ctXor };
+    enum PolyType { ptSubject, ptClip };
+//By far the most widely used winding rules for polygon filling are
+//EvenOdd & NonZero (GDI, GDI+, XLib, OpenGL, Cairo, AGG, Quartz, SVG, Gr32)
+//Others rules include Positive, Negative and ABS_GTR_EQ_TWO (only in OpenGL)
+//see http://glprogramming.com/red/chapter11.html
+    enum PolyFillType { pftEvenOdd, pftNonZero, pftPositive, pftNegative };
+
+#ifdef use_int32
+    typedef int cInt;
+  static cInt const loRange = 0x7FFF;
+  static cInt const hiRange = 0x7FFF;
+#else
+    typedef signed long long cInt;
+    static cInt const loRange = 0x3FFFFFFF;
+    static cInt const hiRange = 0x3FFFFFFFFFFFFFFFLL;
+    typedef signed long long long64;     //used by Int128 class
+    typedef unsigned long long ulong64;
+
+#endif
+
+    struct IntPoint {
+        cInt X;
+        cInt Y;
+#ifdef use_xyz
+        cInt Z;
+  IntPoint(cInt x = 0, cInt y = 0, cInt z = 0): X(x), Y(y), Z(z) {};
+#else
+        IntPoint(cInt x = 0, cInt y = 0): X(x), Y(y) {};
+#endif
+
+        friend inline bool operator== (const IntPoint& a, const IntPoint& b)
+        {
+            return a.X == b.X && a.Y == b.Y;
+        }
+        friend inline bool operator!= (const IntPoint& a, const IntPoint& b)
+        {
+            return a.X != b.X  || a.Y != b.Y;
+        }
+    };
+//------------------------------------------------------------------------------
+
+    typedef std::vector< IntPoint > Path;
+    typedef std::vector< Path > Paths;
+
+    inline Path& operator <<(Path& poly, const IntPoint& p) {poly.push_back(p); return poly;}
+    inline Paths& operator <<(Paths& polys, const Path& p) {polys.push_back(p); return polys;}
+
+    std::ostream& operator <<(std::ostream &s, const IntPoint &p);
+    std::ostream& operator <<(std::ostream &s, const Path &p);
+    std::ostream& operator <<(std::ostream &s, const Paths &p);
+
+    struct DoublePoint
+    {
+        double X;
+        double Y;
+        DoublePoint(double x = 0, double y = 0) : X(x), Y(y) {}
+        DoublePoint(IntPoint ip) : X((double)ip.X), Y((double)ip.Y) {}
+    };
+//------------------------------------------------------------------------------
+
+#ifdef use_xyz
+    typedef void (*ZFillCallback)(IntPoint& e1bot, IntPoint& e1top, IntPoint& e2bot, IntPoint& e2top, IntPoint& pt);
+#endif
+
+    enum InitOptions {ioReverseSolution = 1, ioStrictlySimple = 2, ioPreserveCollinear = 4};
+    enum JoinType {jtSquare, jtRound, jtMiter};
+    enum EndType {etClosedPolygon, etClosedLine, etOpenButt, etOpenSquare, etOpenRound};
+
+    class PolyNode;
+    typedef std::vector< PolyNode* > PolyNodes;
+
+    class PolyNode
+    {
+    public:
+        PolyNode();
+        virtual ~PolyNode(){};
+        Path Contour;
+        PolyNodes Childs;
+        PolyNode* Parent;
+        PolyNode* GetNext() const;
+        bool IsHole() const;
+        bool IsOpen() const;
+        int ChildCount() const;
+    private:
+        //PolyNode& operator =(PolyNode& other);
+        unsigned Index; //node index in Parent.Childs
+        bool m_IsOpen;
+        JoinType m_jointype;
+        EndType m_endtype;
+        PolyNode* GetNextSiblingUp() const;
+        void AddChild(PolyNode& child);
+        friend class Clipper; //to access Index
+        friend class ClipperOffset;
+    };
+
+    class PolyTree: public PolyNode
+    {
+    public:
+        ~PolyTree(){ Clear(); };
+        PolyNode* GetFirst() const;
+        void Clear();
+        int Total() const;
+    private:
+        //PolyTree& operator =(PolyTree& other);
+        PolyNodes AllNodes;
+        friend class Clipper; //to access AllNodes
+    };
+
+    bool Orientation(const Path &poly);
+    double Area(const Path &poly);
+    int PointInPolygon(const IntPoint &pt, const Path &path);
+
+    void SimplifyPolygon(const Path &in_poly, Paths &out_polys, PolyFillType fillType = pftEvenOdd);
+    void SimplifyPolygons(const Paths &in_polys, Paths &out_polys, PolyFillType fillType = pftEvenOdd);
+    void SimplifyPolygons(Paths &polys, PolyFillType fillType = pftEvenOdd);
+
+    void CleanPolygon(const Path& in_poly, Path& out_poly, double distance = 1.415);
+    void CleanPolygon(Path& poly, double distance = 1.415);
+    void CleanPolygons(const Paths& in_polys, Paths& out_polys, double distance = 1.415);
+    void CleanPolygons(Paths& polys, double distance = 1.415);
+
+    void MinkowskiSum(const Path& pattern, const Path& path, Paths& solution, bool pathIsClosed);
+    void MinkowskiSum(const Path& pattern, const Paths& paths, Paths& solution, bool pathIsClosed);
+    void MinkowskiDiff(const Path& poly1, const Path& poly2, Paths& solution);
+
+    void PolyTreeToPaths(const PolyTree& polytree, Paths& paths);
+    void ClosedPathsFromPolyTree(const PolyTree& polytree, Paths& paths);
+    void OpenPathsFromPolyTree(PolyTree& polytree, Paths& paths);
+
+    void ReversePath(Path& p);
+    void ReversePaths(Paths& p);
+
+    struct IntRect { cInt left; cInt top; cInt right; cInt bottom; };
+
+//enums that are used internally ...
+    enum EdgeSide { esLeft = 1, esRight = 2};
+
+//forward declarations (for stuff used internally) ...
+    struct TEdge;
+    struct IntersectNode;
+    struct LocalMinimum;
+    struct OutPt;
+    struct OutRec;
+    struct Join;
+
+    typedef std::vector < OutRec* > PolyOutList;
+    typedef std::vector < TEdge* > EdgeList;
+    typedef std::vector < Join* > JoinList;
+    typedef std::vector < IntersectNode* > IntersectList;
+
+//------------------------------------------------------------------------------
+
+//ClipperBase is the ancestor to the Clipper class. It should not be
+//instantiated directly. This class simply abstracts the conversion of sets of
+//polygon coordinates into edge objects that are stored in a LocalMinima list.
+    class ClipperBase
+    {
+    public:
+        ClipperBase();
+        virtual ~ClipperBase();
+        virtual bool AddPath(const Path &pg, PolyType PolyTyp, bool Closed);
+        bool AddPaths(const Paths &ppg, PolyType PolyTyp, bool Closed);
+        virtual void Clear();
+        IntRect GetBounds();
+        bool PreserveCollinear() {return m_PreserveCollinear;};
+        void PreserveCollinear(bool value) {m_PreserveCollinear = value;};
+    protected:
+        void DisposeLocalMinimaList();
+        TEdge* AddBoundsToLML(TEdge *e, bool IsClosed);
+        virtual void Reset();
+        TEdge* ProcessBound(TEdge* E, bool IsClockwise);
+        void InsertScanbeam(const cInt Y);
+        bool PopScanbeam(cInt &Y);
+        bool LocalMinimaPending();
+        bool PopLocalMinima(cInt Y, const LocalMinimum *&locMin);
+        OutRec* CreateOutRec();
+        void DisposeAllOutRecs();
+        void DisposeOutRec(PolyOutList::size_type index);
+        void SwapPositionsInAEL(TEdge *edge1, TEdge *edge2);
+        void DeleteFromAEL(TEdge *e);
+        void UpdateEdgeIntoAEL(TEdge *&e);
+
+        typedef std::vector<LocalMinimum> MinimaList;
+        MinimaList::iterator m_CurrentLM;
+        MinimaList           m_MinimaList;
+
+        bool              m_UseFullRange;
+        EdgeList          m_edges;
+        bool              m_PreserveCollinear;
+        bool              m_HasOpenPaths;
+        PolyOutList       m_PolyOuts;
+        TEdge           *m_ActiveEdges;
+
+        typedef std::priority_queue<cInt> ScanbeamList;
+        ScanbeamList     m_Scanbeam;
+    };
+//------------------------------------------------------------------------------
+
+    class Clipper : public virtual ClipperBase
+    {
+    public:
+        Clipper(int initOptions = 0);
+        bool Execute(ClipType clipType,
+                     Paths &solution,
+                     PolyFillType fillType = pftEvenOdd);
+        bool Execute(ClipType clipType,
+                     Paths &solution,
+                     PolyFillType subjFillType,
+                     PolyFillType clipFillType);
+        bool Execute(ClipType clipType,
+                     PolyTree &polytree,
+                     PolyFillType fillType = pftEvenOdd);
+        bool Execute(ClipType clipType,
+                     PolyTree &polytree,
+                     PolyFillType subjFillType,
+                     PolyFillType clipFillType);
+        bool ReverseSolution() { return m_ReverseOutput; };
+        void ReverseSolution(bool value) {m_ReverseOutput = value;};
+        bool StrictlySimple() {return m_StrictSimple;};
+        void StrictlySimple(bool value) {m_StrictSimple = value;};
+        //set the callback function for z value filling on intersections (otherwise Z is 0)
+#ifdef use_xyz
+        void ZFillFunction(ZFillCallback zFillFunc);
+#endif
+    protected:
+        virtual bool ExecuteInternal();
+    private:
+        JoinList         m_Joins;
+        JoinList         m_GhostJoins;
+        IntersectList    m_IntersectList;
+        ClipType         m_ClipType;
+        typedef std::list<cInt> MaximaList;
+        MaximaList       m_Maxima;
+        TEdge           *m_SortedEdges;
+        bool             m_ExecuteLocked;
+        PolyFillType     m_ClipFillType;
+        PolyFillType     m_SubjFillType;
+        bool             m_ReverseOutput;
+        bool             m_UsingPolyTree;
+        bool             m_StrictSimple;
+#ifdef use_xyz
+        ZFillCallback   m_ZFill; //custom callback
+#endif
+        void SetWindingCount(TEdge& edge);
+        bool IsEvenOddFillType(const TEdge& edge) const;
+        bool IsEvenOddAltFillType(const TEdge& edge) const;
+        void InsertLocalMinimaIntoAEL(const cInt botY);
+        void InsertEdgeIntoAEL(TEdge *edge, TEdge* startEdge);
+        void AddEdgeToSEL(TEdge *edge);
+        bool PopEdgeFromSEL(TEdge *&edge);
+        void CopyAELToSEL();
+        void DeleteFromSEL(TEdge *e);
+        void SwapPositionsInSEL(TEdge *edge1, TEdge *edge2);
+        bool IsContributing(const TEdge& edge) const;
+        bool IsTopHorz(const cInt XPos);
+        void DoMaxima(TEdge *e);
+        void ProcessHorizontals();
+        void ProcessHorizontal(TEdge *horzEdge);
+        void AddLocalMaxPoly(TEdge *e1, TEdge *e2, const IntPoint &pt);
+        OutPt* AddLocalMinPoly(TEdge *e1, TEdge *e2, const IntPoint &pt);
+        OutRec* GetOutRec(int idx);
+        void AppendPolygon(TEdge *e1, TEdge *e2);
+        void IntersectEdges(TEdge *e1, TEdge *e2, IntPoint &pt);
+        OutPt* AddOutPt(TEdge *e, const IntPoint &pt);
+        OutPt* GetLastOutPt(TEdge *e);
+        bool ProcessIntersections(const cInt topY);
+        void BuildIntersectList(const cInt topY);
+        void ProcessIntersectList();
+        void ProcessEdgesAtTopOfScanbeam(const cInt topY);
+        void BuildResult(Paths& polys);
+        void BuildResult2(PolyTree& polytree);
+        void SetHoleState(TEdge *e, OutRec *outrec);
+        void DisposeIntersectNodes();
+        bool FixupIntersectionOrder();
+        void FixupOutPolygon(OutRec &outrec);
+        void FixupOutPolyline(OutRec &outrec);
+        bool IsHole(TEdge *e);
+        bool FindOwnerFromSplitRecs(OutRec &outRec, OutRec *&currOrfl);
+        void FixHoleLinkage(OutRec &outrec);
+        void AddJoin(OutPt *op1, OutPt *op2, const IntPoint offPt);
+        void ClearJoins();
+        void ClearGhostJoins();
+        void AddGhostJoin(OutPt *op, const IntPoint offPt);
+        bool JoinPoints(Join *j, OutRec* outRec1, OutRec* outRec2);
+        void JoinCommonEdges();
+        void DoSimplePolygons();
+        void FixupFirstLefts1(OutRec* OldOutRec, OutRec* NewOutRec);
+        void FixupFirstLefts2(OutRec* InnerOutRec, OutRec* OuterOutRec);
+        void FixupFirstLefts3(OutRec* OldOutRec, OutRec* NewOutRec);
+#ifdef use_xyz
+        void SetZ(IntPoint& pt, TEdge& e1, TEdge& e2);
+#endif
+    };
+//------------------------------------------------------------------------------
+
+    class ClipperOffset
+    {
+    public:
+        ClipperOffset(double miterLimit = 2.0, double roundPrecision = 0.25);
+        ~ClipperOffset();
+        void AddPath(const Path& path, JoinType joinType, EndType endType);
+        void AddPaths(const Paths& paths, JoinType joinType, EndType endType);
+        void Execute(Paths& solution, double delta);
+        void Execute(PolyTree& solution, double delta);
+        void Clear();
+        double MiterLimit;
+        double ArcTolerance;
+    private:
+        Paths m_destPolys;
+        Path m_srcPoly;
+        Path m_destPoly;
+        std::vector<DoublePoint> m_normals;
+        double m_delta, m_sinA, m_sin, m_cos;
+        double m_miterLim, m_StepsPerRad;
+        IntPoint m_lowest;
+        PolyNode m_polyNodes;
+
+        void FixOrientations();
+        void DoOffset(double delta);
+        void OffsetPoint(int j, int& k, JoinType jointype);
+        void DoSquare(int j, int k);
+        void DoMiter(int j, int k, double r);
+        void DoRound(int j, int k);
+    };
+//------------------------------------------------------------------------------
+
+    class clipperException : public std::exception
+    {
+    public:
+        clipperException(const char* description): m_descr(description) {}
+        virtual ~clipperException() throw() {}
+        virtual const char* what() const throw() {return m_descr.c_str();}
+    private:
+        std::string m_descr;
+    };
+//------------------------------------------------------------------------------
+
+} //ClipperLib namespace
+
+#endif //clipper_hpp
+
+

3rdPart/include/OcrLiteOnnx/getopt.h

@@ -0,0 +1,47 @@
+/*
+ * getopt - POSIX like getopt for Windows console Application
+ *
+ * win-c - Windows Console Library
+ * Copyright (c) 2015 Koji Takami
+ * Released under the MIT license
+ * https://github.com/takamin/win-c/blob/master/LICENSE
+ */
+#ifndef _GETOPT_H_
+#define _GETOPT_H_
+
+#ifndef __CLIB__
+
+#ifdef __cplusplus
+extern "C" {
+#endif // __cplusplus
+
+int getopt(int argc, char *const argv[],
+           const char *optstring);
+
+extern char *optarg;
+extern int optind, opterr, optopt;
+
+#define no_argument 0
+#define required_argument 1
+#define optional_argument 2
+
+struct option {
+    const char *name;
+    int has_arg;
+    int *flag;
+    int val;
+};
+
+int getopt_long(int argc, char *const argv[],
+                const char *optstring,
+                const struct option *longopts, int *longindex);
+/****************************************************************************
+    int getopt_long_only(int argc, char* const argv[],
+            const char* optstring,
+            const struct option* longopts, int* longindex);
+****************************************************************************/
+#ifdef __cplusplus
+}
+#endif
+#endif // __cplusplus
+#endif // _GETOPT_H_

3rdPart/include/OcrLiteOnnx/main.h

@@ -0,0 +1,56 @@
+#ifndef __MAIN_H__
+#define __MAIN_H__
+#ifndef __CLIB__
+
+#include "getopt.h"
+
+static const struct option long_options[] = {
+        {"models",         required_argument, NULL, 'd'},
+        {"det",            required_argument, NULL, '1'},
+        {"cls",            required_argument, NULL, '2'},
+        {"rec",            required_argument, NULL, '3'},
+        {"keys",           required_argument, NULL, '4'},
+        {"image",          required_argument, NULL, 'i'},
+        {"numThread",      required_argument, NULL, 't'},
+        {"padding",        required_argument, NULL, 'p'},
+        {"maxSideLen",     required_argument, NULL, 's'},
+        {"boxScoreThresh", required_argument, NULL, 'b'},
+        {"boxThresh",      required_argument, NULL, 'o'},
+        {"unClipRatio",    required_argument, NULL, 'u'},
+        {"doAngle",        required_argument, NULL, 'a'},
+        {"mostAngle",      required_argument, NULL, 'A'},
+        {"version",        no_argument,       NULL, 'v'},
+        {"help",           no_argument,       NULL, 'h'},
+        {"loopCount",      required_argument, NULL, 'l'},
+        {NULL,             no_argument,       NULL, 0}
+};
+
+const char *usageMsg = "(-d --models) (-1 --det) (-2 --cls) (-3 --rec) (-4 --keys) (-i --image)\n"\
+                       "[-t --numThread] [-p --padding] [-s --maxSideLen]\n" \
+                       "[-b --boxScoreThresh] [-o --boxThresh] [-u --unClipRatio]\n" \
+                       "[-a --noAngle] [-A --mostAngle]\n\n";
+
+const char *requiredMsg = "-d --models: models directory.\n" \
+                          "-1 --det: model file name of det.\n" \
+                          "-2 --cls: model file name of cls.\n" \
+                          "-3 --rec: model file name of rec.\n" \
+						  "-4 --keys: keys file name.\n" \
+                          "-i --image: path of target image.\n\n";
+
+const char *optionalMsg = "-t --numThread: value of numThread(int), default: 4\n" \
+                          "-p --padding: value of padding(int), default: 50\n" \
+                          "-s --maxSideLen: Long side of picture for resize(int), default: 1024\n" \
+                          "-b --boxScoreThresh: value of boxScoreThresh(float), default: 0.6\n" \
+                          "-o --boxThresh: value of boxThresh(float), default: 0.3\n" \
+                          "-u --unClipRatio: value of unClipRatio(float), default: 2.0\n" \
+                          "-a --doAngle: Enable(1)/Disable(0) Angle Net, default: Enable\n" \
+                          "-A --mostAngle: Enable(1)/Disable(0) Most Possible AngleIndex, default: Enable\n\n";
+
+const char *otherMsg = "-v --version: show version\n" \
+                       "-h --help: print this help\n\n";
+
+const char *example1Msg = "Example1: %s --models models --det det.onnx --cls cls.onnx --rec rec.onnx --keys keys.txt --image 1.jpg\n";
+const char *example2Msg = "Example2: %s -d models -1 det.onnx -2 cls.onnx -3 rec.onnx -4 keys.txt -i 1.jpg -t 4 -p 50 -s 0 -b 0.6 -o 0.3 -u 2.0 -a 1 -A 1\n";
+
+#endif
+#endif //__MAIN_H__

3rdPart/include/OcrLiteOnnx/version.h

@@ -0,0 +1,6 @@
+#ifndef __OCR_VERSION_H__
+#define __OCR_VERSION_H__
+
+#define VERSION "1.5.1.20210128"
+
+#endif //__OCR_VERSION_H__

3rdPart/include/meojson/json.h

@@ -0,0 +1,7 @@
+#pragma once
+
+#include "json_parser.h"
+#include "json_value.h"
+#include "json_object.h"
+#include "json_array.h"
+#include "json_exception.h"

3rdPart/include/meojson/json_array.h

@@ -0,0 +1,93 @@
+#pragma once
+
+#include <string>
+#include <vector>
+#include <initializer_list>
+
+namespace json
+{
+    class value;
+
+    class array
+    {
+    public:
+        using raw_array = std::vector<value>;
+        using iterator = raw_array::iterator;
+        using const_iterator = raw_array::const_iterator;
+        using reverse_iterator = raw_array::reverse_iterator;
+        using const_reverse_iterator = raw_array::const_reverse_iterator;
+
+        array() = default;
+        array(const array& rhs) = default;
+        array(array&& rhs) noexcept = default;
+        array(const raw_array& arr);
+        array(raw_array&& arr) noexcept;
+        array(std::initializer_list<raw_array::value_type> init_list);
+        template<typename EleType>
+        array(std::vector<EleType> vec) {
+            static_assert(
+                std::is_constructible<json::value, EleType>::value,
+                "Parameter can't be used to construct a json::value");
+            for (auto&& ele : vec) {
+                _array_data.emplace_back(std::move(ele));
+            }
+        }
+
+        ~array() noexcept = default;
+
+        bool empty() const noexcept { return _array_data.empty(); }
+        size_t size() const noexcept { return _array_data.size(); }
+        bool exist(size_t pos) const { return _array_data.size() < pos; }
+        const value& at(size_t pos) const;
+        const std::string to_string() const;
+        const std::string format(std::string shift_str = "    ", size_t basic_shift_count = 0) const;
+
+        const bool get(size_t pos, bool default_value) const;
+        const int get(size_t pos, int default_value) const;
+        const long get(size_t pos, long default_value) const;
+        const unsigned long get(size_t pos, unsigned default_value) const;
+        const long long get(size_t pos, long long default_value) const;
+        const unsigned long long get(size_t pos, unsigned long long default_value) const;
+        const float get(size_t pos, float default_value) const;
+        const double get(size_t pos, double default_value) const;
+        const long double get(size_t pos, long double default_value) const;
+        const std::string get(size_t pos, std::string default_value) const;
+        const std::string get(size_t pos, const char* default_value) const;
+
+        template <typename... Args>
+        decltype(auto) emplace_back(Args &&... args)
+        {
+            static_assert(
+                std::is_constructible<raw_array::value_type, Args...>::value,
+                "Parameter can't be used to construct a raw_array::value_type");
+            return _array_data.emplace_back(std::forward<Args>(args)...);
+        }
+
+        void clear() noexcept;
+        // void earse(size_t pos);
+
+        iterator begin() noexcept;
+        iterator end() noexcept;
+        const_iterator cbegin() const noexcept;
+        const_iterator cend() const noexcept;
+
+        reverse_iterator rbegin() noexcept;
+        reverse_iterator rend() noexcept;
+        const_reverse_iterator crbegin() const noexcept;
+        const_reverse_iterator crend() const noexcept;
+
+        const value& operator[](size_t pos) const;
+        value& operator[](size_t pos);
+
+        array& operator=(const array&) = default;
+        array& operator=(array&&) noexcept = default;
+
+        // const raw_array &raw_data() const;
+
+    private:
+        raw_array _array_data;
+    };
+
+    std::ostream& operator<<(std::ostream& out, const array& arr);
+
+} // namespace json

3rdPart/include/meojson/json_aux.h

@@ -0,0 +1,97 @@
+#pragma once
+
+#include <string>
+#include "json_value.h"
+
+namespace json
+{
+	static std::string unescape_string(std::string&& str)
+	{
+		std::string replace_str;
+		std::string escape_str = std::move(str);
+
+		for (size_t pos = 0; pos < escape_str.size(); ++pos)
+		{
+			switch (escape_str[pos]) {
+			case '\"':
+				replace_str = R"(\")";
+				break;
+			case '\\':
+				replace_str = R"(\\)";
+				break;
+			case '\b':
+				replace_str = R"(\b)";
+				break;
+			case '\f':
+				replace_str = R"(\f)";
+				break;
+			case '\n':
+				replace_str = R"(\n)";
+				break;
+			case '\r':
+				replace_str = R"(\r)";
+				break;
+			case '\t':
+				replace_str = R"(\t)";
+				break;
+			default:
+				continue;
+				break;
+			}
+			escape_str.replace(pos, 1, replace_str);
+			++pos;
+		}
+		return escape_str;
+	}
+
+	static std::string unescape_string(const std::string& str)
+	{
+		return unescape_string(std::string(str));
+	}
+
+	static std::string escape_string(std::string&& str)
+	{
+		std::string escape_str = std::move(str);
+
+		for (size_t pos = 0; pos + 1 < escape_str.size(); ++pos)
+		{
+			if (escape_str[pos] != '\\') {
+				continue;
+			}
+			std::string replace_str;
+			switch (escape_str[pos+1]) {
+			case '"':
+				replace_str = "\"";
+				break;
+			case '\\':
+				replace_str = "\\";
+				break;
+			case 'b':
+				replace_str = "\b";
+				break;
+			case 'f':
+				replace_str = "\f";
+				break;
+			case 'n':
+				replace_str = "\n";
+				break;
+			case 'r':
+				replace_str = "\r";
+				break;
+			case 't':
+				replace_str = "\r";
+				break;
+			default:
+				return std::string();
+				break;
+			}
+			escape_str.replace(pos, 2, replace_str);
+		}
+		return escape_str;
+	}
+
+	static std::string escape_string(const std::string& str)
+	{
+		return escape_string(std::string(str));
+	}
+}

3rdPart/include/meojson/json_exception.h

@@ -0,0 +1,27 @@
+#pragma once
+
+#include <exception>
+#include <string>
+
+namespace json
+{
+    class exception : public std::exception
+    {
+    public:
+        exception() = default;
+        exception(const std::string& msg);
+
+        exception(const exception&) = default;
+        exception& operator=(const exception&) = default;
+        exception(exception&&) = default;
+        exception& operator=(exception&&) = default;
+
+        virtual ~exception() noexcept override = default;
+
+        virtual const char* what() const noexcept override;
+
+    private:
+        std::string m_msg;
+    };
+
+} // namespace json

3rdPart/include/meojson/json_object.h

@@ -0,0 +1,77 @@
+#pragma once
+
+#include <string>
+#include <unordered_map>
+#include <initializer_list>
+
+#include "json_value.h"
+
+namespace json
+{
+	class object
+	{
+	public:
+		using raw_object = std::unordered_map<std::string, value>;
+		using iterator = raw_object::iterator;
+		using const_iterator = raw_object::const_iterator;
+
+		object() = default;
+		object(const object& rhs) = default;
+		object(object&& rhs) = default;
+		object(const raw_object& raw_obj);
+		object(raw_object&& raw_obj);
+		object(std::initializer_list<raw_object::value_type> init_list);
+
+		~object() = default;
+
+		bool empty() const noexcept { return _object_data.empty(); }
+		size_t size() const noexcept { return _object_data.size(); }
+		bool exist(const std::string& key) const { return _object_data.find(key) != _object_data.cend(); }
+		const value& at(const std::string& key) const;
+		const std::string to_string() const;
+		const std::string format(std::string shift_str = "    ", size_t basic_shift_count = 0) const;
+
+		const bool get(const std::string& key, bool default_value) const;
+		const int get(const std::string& key, int default_value) const;
+		const long get(const std::string& key, long default_value) const;
+		const unsigned long get(const std::string& key, unsigned default_value) const;
+		const long long get(const std::string& key, long long default_value) const;
+		const unsigned long long get(const std::string& key, unsigned long long default_value) const;
+		const float get(const std::string& key, float default_value) const;
+		const double get(const std::string& key, double default_value) const;
+		const long double get(const std::string& key, long double default_value) const;
+		const std::string get(const std::string& key, std::string default_value) const;
+		const std::string get(const std::string& key, const char* default_value) const;
+
+		template <typename... Args>
+		decltype(auto) emplace(Args &&... args)
+		{
+			static_assert(
+				std::is_constructible<raw_object::value_type, Args...>::value,
+				"Parameter can't be used to construct a raw_object::value_type");
+			return _object_data.emplace(std::forward<Args>(args)...);
+		}
+
+		void clear() noexcept;
+		bool earse(const std::string& key);
+
+		iterator begin() noexcept;
+		iterator end() noexcept;
+		const_iterator cbegin() const noexcept;
+		const_iterator cend() const noexcept;
+
+		value& operator[](const std::string& key);
+		value& operator[](std::string&& key);
+
+		object& operator=(const object&) = default;
+		object& operator=(object&&) = default;
+
+		// const raw_object &raw_data() const;
+
+	private:
+		raw_object _object_data;
+	};
+
+	std::ostream& operator<<(std::ostream& out, const json::object& obj);
+
+} // namespace json

3rdPart/include/meojson/json_parser.h

@@ -0,0 +1,50 @@
+#pragma once
+
+#include <string>
+#include <optional>
+
+namespace json
+{
+    class value;
+    class object;
+    class array;
+    enum class value_type : char;
+
+    class parser
+    {
+    public:
+        ~parser() noexcept = default;
+
+        static std::optional<value> parse(const std::string& content);
+
+    private:
+        parser(
+            const std::string::const_iterator& cbegin,
+            const std::string::const_iterator& cend) noexcept
+            : _cur(cbegin), _end(cend) {}
+
+        std::optional<value> parse();
+        value parse_value();
+
+        value parse_null();
+        value parse_boolean();
+        value parse_number();
+        // parse and return a json::value whose type is value_type::String
+        value parse_string();
+        value parse_array();
+        value parse_object();
+
+        // parse and return a std::string
+        std::optional<std::string> parse_stdstring();
+
+        bool skip_whitespace() noexcept;
+        bool skip_digit() noexcept;
+
+    private:
+        std::string::const_iterator _cur;
+        std::string::const_iterator _end;
+    };
+
+    std::optional<value> parse(const std::string& content);
+
+} // namespace json

3rdPart/include/meojson/json_value.h

@@ -0,0 +1,160 @@
+#pragma once
+
+#include <string>
+#include <ostream>
+// #include <iostream>
+#include <memory>
+
+namespace json
+{
+    enum class value_type : char
+    {
+        Invalid,
+        Null,
+        Boolean,
+        String,
+        Number,
+        Array,
+        Object,
+        NUM_T
+    };
+
+    class array;
+    class object;
+
+    class value
+    {
+        using unique_array = std::unique_ptr<array>;
+        using unique_object = std::unique_ptr<object>;
+
+    public:
+        value();
+        value(const value& rhs);
+        value(value&& rhs) noexcept;
+
+        value(bool b);
+
+        value(int num);
+        value(unsigned num);
+        value(long num);
+        value(unsigned long num);
+        value(long long num);
+        value(unsigned long long num);
+        value(float num);
+        value(double num);
+        value(long double num);
+
+        value(const char* str);
+        value(const std::string& str);
+        value(std::string&& str);
+
+        value(const array& arr);
+        value(array&& arr);
+        // value(std::initializer_list<value> init_list); // for array
+
+        value(const object& obj);
+        value(object&& obj);
+        // error: conversion from ‘<brace-enclosed initializer list>’ to ‘json::value’ is ambiguous
+        // value(std::initializer_list<std::pair<std::string, value>> init_list); // for object
+
+        // Constructed from raw data
+        template <typename... Args>
+        value(value_type type, Args &&...args)
+            : _type(type),
+            _raw_data(std::forward<Args>(args)...)
+        {
+            static_assert(
+                std::is_constructible<std::string, Args...>::value,
+                "Parameter can't be used to construct a std::string");
+        }
+
+        // Prohibit conversion of other types to value
+        template <typename T>
+        value(T) = delete;
+
+        ~value();
+
+        bool valid() const noexcept { return _type != value_type::Invalid ? true : false; }
+        bool empty() const noexcept { return (_type == value_type::Null && _raw_data.compare("null") == 0) ? true : false; }
+        bool is_null() const noexcept { return empty(); }
+        bool is_number() const noexcept { return _type == value_type::Number; }
+        bool is_boolean() const noexcept { return _type == value_type::Boolean; }
+        bool is_string() const noexcept { return _type == value_type::String; }
+        bool is_array() const noexcept { return _type == value_type::Array; }
+        bool is_object() const noexcept { return _type == value_type::Object; }
+        bool exist(const std::string& key) const;
+        bool exist(size_t pos) const;
+        value_type type() const noexcept { return _type; }
+        const value& at(size_t pos) const;
+        const value& at(const std::string& key) const;
+
+        template <typename Type>
+        decltype(auto) get(const std::string& key, Type default_value) const
+        {
+            return is_object() ? as_object().get(key, default_value) : default_value;
+        }
+        template <typename Type>
+        decltype(auto) get(size_t pos, Type default_value) const
+        {
+            return is_array() ? as_array().get(pos, default_value) : default_value;
+        }
+
+        const bool as_boolean() const;
+        const int as_integer() const;
+        // const unsigned as_unsigned() const;
+        const long as_long() const;
+        const unsigned long as_unsigned_long() const;
+        const long long as_long_long() const;
+        const unsigned long long as_unsigned_long_long() const;
+        const float as_float() const;
+        const double as_double() const;
+        const long double as_long_double() const;
+        const std::string as_string() const;
+        const array& as_array() const;
+        const object& as_object() const;
+
+        array& as_array();
+        object& as_object();
+
+        // return raw string
+        const std::string to_string() const;
+        const std::string format(std::string shift_str = "    ", size_t basic_shift_count = 0) const;
+
+        value& operator=(const value& rhs);
+        value& operator=(value&&) noexcept;
+
+        const value& operator[](size_t pos) const;
+        value& operator[](size_t pos);
+        value& operator[](const std::string& key);
+        value& operator[](std::string&& key);
+        //explicit operator bool() const noexcept { return valid(); }
+
+        explicit operator bool() const { return as_boolean(); }
+        explicit operator int() const { return as_integer(); }
+        explicit operator long() const { return as_long(); }
+        explicit operator unsigned long() const { return as_unsigned_long(); }
+        explicit operator long long() const { return as_long_long(); }
+        explicit operator unsigned long long() const { return as_unsigned_long_long(); }
+        explicit operator float() const { return as_float(); }
+        explicit operator double() const { return as_double(); }
+        explicit operator long double() const { return as_long_double(); }
+        explicit operator std::string() const { return as_string(); }
+
+    private:
+        template <typename T>
+        static std::unique_ptr<T> copy_unique_ptr(const std::unique_ptr<T>& t)
+        {
+            return t == nullptr ? nullptr : std::make_unique<T>(*t);
+        }
+
+        value_type _type = value_type::Null;
+        std::string _raw_data = "null"; // If the value_type is Object or Array, the _raw_data will be a empty string.
+        unique_array _array_ptr;
+        unique_object _object_ptr;
+    };
+
+    const value invalid_value();
+    std::ostream& operator<<(std::ostream& out, const value& val);
+    // std::istream &operator>>(std::istream &in, value &val);
+
+} // namespace json

3rdPart/include/opencv/cv.h

@@ -0,0 +1,73 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_CV_H
+#define OPENCV_OLD_CV_H
+
+#if defined(_MSC_VER)
+    #define CV_DO_PRAGMA(x) __pragma(x)
+    #define __CVSTR2__(x) #x
+    #define __CVSTR1__(x) __CVSTR2__(x)
+    #define __CVMSVCLOC__ __FILE__ "("__CVSTR1__(__LINE__)") : "
+    #define CV_MSG_PRAGMA(_msg) CV_DO_PRAGMA(message (__CVMSVCLOC__ _msg))
+#elif defined(__GNUC__)
+    #define CV_DO_PRAGMA(x) _Pragma (#x)
+    #define CV_MSG_PRAGMA(_msg) CV_DO_PRAGMA(message (_msg))
+#else
+    #define CV_DO_PRAGMA(x)
+    #define CV_MSG_PRAGMA(_msg)
+#endif
+#define CV_WARNING(x) CV_MSG_PRAGMA("Warning: " #x)
+
+//CV_WARNING("This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module")
+
+#include "opencv2/core/core_c.h"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/photo/photo_c.h"
+#include "opencv2/video/tracking_c.h"
+#include "opencv2/objdetect/objdetect_c.h"
+
+#if !defined(CV_IMPL)
+#define CV_IMPL extern "C"
+#endif //CV_IMPL
+
+#endif // __OPENCV_OLD_CV_H_

3rdPart/include/opencv/cv.hpp

@@ -0,0 +1,60 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_CV_HPP
+#define OPENCV_OLD_CV_HPP
+
+//#if defined(__GNUC__)
+//#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
+//#endif
+
+#include "cv.h"
+#include "opencv2/core.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/photo.hpp"
+#include "opencv2/video.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/objdetect.hpp"
+
+#endif

3rdPart/include/opencv/cvaux.h

@@ -0,0 +1,57 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                        Intel License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of Intel Corporation may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_AUX_H
+#define OPENCV_OLD_AUX_H
+
+//#if defined(__GNUC__)
+//#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
+//#endif
+
+#include "opencv2/core/core_c.h"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/photo/photo_c.h"
+#include "opencv2/video/tracking_c.h"
+#include "opencv2/objdetect/objdetect_c.h"
+
+#endif
+
+/* End of file. */

3rdPart/include/opencv/cvaux.hpp

@@ -0,0 +1,52 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                        Intel License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of Intel Corporation may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_AUX_HPP
+#define OPENCV_OLD_AUX_HPP
+
+//#if defined(__GNUC__)
+//#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
+//#endif
+
+#include "cvaux.h"
+#include "opencv2/core/utility.hpp"
+
+#endif

3rdPart/include/opencv/cvwimage.h

@@ -0,0 +1,46 @@
+///////////////////////////////////////////////////////////////////////////////
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to
+//  this license.  If you do not agree to this license, do not download,
+//  install, copy or use the software.
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2008, Google, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//  * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//  * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//  * The name of Intel Corporation or contributors may not be used to endorse
+//     or promote products derived from this software without specific
+//     prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is"
+// and any express or implied warranties, including, but not limited to, the
+// implied warranties of merchantability and fitness for a particular purpose
+// are disclaimed. In no event shall the Intel Corporation or contributors be
+// liable for any direct, indirect, incidental, special, exemplary, or
+// consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+
+
+#ifndef OPENCV_OLD_WIMAGE_HPP
+#define OPENCV_OLD_WIMAGE_HPP
+
+#include "opencv2/core/wimage.hpp"
+
+#endif

3rdPart/include/opencv/cxcore.h

@@ -0,0 +1,52 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_CXCORE_H
+#define OPENCV_OLD_CXCORE_H
+
+//#if defined(__GNUC__)
+//#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
+//#endif
+
+#include "opencv2/core/core_c.h"
+
+#endif

3rdPart/include/opencv/cxcore.hpp

@@ -0,0 +1,53 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_CXCORE_HPP
+#define OPENCV_OLD_CXCORE_HPP
+
+//#if defined(__GNUC__)
+//#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
+//#endif
+
+#include "cxcore.h"
+#include "opencv2/core.hpp"
+
+#endif

3rdPart/include/opencv/cxeigen.hpp

@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_EIGEN_HPP
+#define OPENCV_OLD_EIGEN_HPP
+
+#include "opencv2/core/eigen.hpp"
+
+#endif

3rdPart/include/opencv/cxmisc.h

@@ -0,0 +1,8 @@
+#ifndef OPENCV_OLD_CXMISC_H
+#define OPENCV_OLD_CXMISC_H
+
+#ifdef __cplusplus
+#  include "opencv2/core/utility.hpp"
+#endif
+
+#endif

3rdPart/include/opencv/highgui.h

@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                        Intel License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of Intel Corporation may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_HIGHGUI_H
+#define OPENCV_OLD_HIGHGUI_H
+
+#include "opencv2/core/core_c.h"
+#include "opencv2/highgui/highgui_c.h"
+
+#endif

3rdPart/include/opencv/ml.h

@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                        Intel License Agreement
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of Intel Corporation may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_OLD_ML_H
+#define OPENCV_OLD_ML_H
+
+#include "opencv2/core/core_c.h"
+#include "opencv2/ml.hpp"
+
+#endif

3rdPart/include/opencv2/calib3d/calib3d.hpp

@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __OPENCV_BUILD
+#error this is a compatibility header which should not be used inside the OpenCV library
+#endif
+
+#include "opencv2/calib3d.hpp"

3rdPart/include/opencv2/calib3d/calib3d_c.h

@@ -0,0 +1,427 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CALIB3D_C_H
+#define OPENCV_CALIB3D_C_H
+
+#include "opencv2/core/core_c.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/** @addtogroup calib3d_c
+  @{
+  */
+
+/****************************************************************************************\
+*                      Camera Calibration, Pose Estimation and Stereo                    *
+\****************************************************************************************/
+
+typedef struct CvPOSITObject CvPOSITObject;
+
+/* Allocates and initializes CvPOSITObject structure before doing cvPOSIT */
+CVAPI(CvPOSITObject*)  cvCreatePOSITObject( CvPoint3D32f* points, int point_count );
+
+
+/* Runs POSIT (POSe from ITeration) algorithm for determining 3d position of
+   an object given its model and projection in a weak-perspective case */
+CVAPI(void)  cvPOSIT(  CvPOSITObject* posit_object, CvPoint2D32f* image_points,
+                       double focal_length, CvTermCriteria criteria,
+                       float* rotation_matrix, float* translation_vector);
+
+/* Releases CvPOSITObject structure */
+CVAPI(void)  cvReleasePOSITObject( CvPOSITObject**  posit_object );
+
+/* updates the number of RANSAC iterations */
+CVAPI(int) cvRANSACUpdateNumIters( double p, double err_prob,
+                                   int model_points, int max_iters );
+
+CVAPI(void) cvConvertPointsHomogeneous( const CvMat* src, CvMat* dst );
+
+/* Calculates fundamental matrix given a set of corresponding points */
+#define CV_FM_7POINT 1
+#define CV_FM_8POINT 2
+
+#define CV_LMEDS 4
+#define CV_RANSAC 8
+
+#define CV_FM_LMEDS_ONLY  CV_LMEDS
+#define CV_FM_RANSAC_ONLY CV_RANSAC
+#define CV_FM_LMEDS CV_LMEDS
+#define CV_FM_RANSAC CV_RANSAC
+
+enum
+{
+    CV_ITERATIVE = 0,
+    CV_EPNP = 1, // F.Moreno-Noguer, V.Lepetit and P.Fua "EPnP: Efficient Perspective-n-Point Camera Pose Estimation"
+    CV_P3P = 2, // X.S. Gao, X.-R. Hou, J. Tang, H.-F. Chang; "Complete Solution Classification for the Perspective-Three-Point Problem"
+    CV_DLS = 3 // Joel A. Hesch and Stergios I. Roumeliotis. "A Direct Least-Squares (DLS) Method for PnP"
+};
+
+CVAPI(int) cvFindFundamentalMat( const CvMat* points1, const CvMat* points2,
+                                 CvMat* fundamental_matrix,
+                                 int method CV_DEFAULT(CV_FM_RANSAC),
+                                 double param1 CV_DEFAULT(3.), double param2 CV_DEFAULT(0.99),
+                                 CvMat* status CV_DEFAULT(NULL) );
+
+/* For each input point on one of images
+   computes parameters of the corresponding
+   epipolar line on the other image */
+CVAPI(void) cvComputeCorrespondEpilines( const CvMat* points,
+                                         int which_image,
+                                         const CvMat* fundamental_matrix,
+                                         CvMat* correspondent_lines );
+
+/* Triangulation functions */
+
+CVAPI(void) cvTriangulatePoints(CvMat* projMatr1, CvMat* projMatr2,
+                                CvMat* projPoints1, CvMat* projPoints2,
+                                CvMat* points4D);
+
+CVAPI(void) cvCorrectMatches(CvMat* F, CvMat* points1, CvMat* points2,
+                             CvMat* new_points1, CvMat* new_points2);
+
+
+/* Computes the optimal new camera matrix according to the free scaling parameter alpha:
+   alpha=0 - only valid pixels will be retained in the undistorted image
+   alpha=1 - all the source image pixels will be retained in the undistorted image
+*/
+CVAPI(void) cvGetOptimalNewCameraMatrix( const CvMat* camera_matrix,
+                                         const CvMat* dist_coeffs,
+                                         CvSize image_size, double alpha,
+                                         CvMat* new_camera_matrix,
+                                         CvSize new_imag_size CV_DEFAULT(cvSize(0,0)),
+                                         CvRect* valid_pixel_ROI CV_DEFAULT(0),
+                                         int center_principal_point CV_DEFAULT(0));
+
+/* Converts rotation vector to rotation matrix or vice versa */
+CVAPI(int) cvRodrigues2( const CvMat* src, CvMat* dst,
+                         CvMat* jacobian CV_DEFAULT(0) );
+
+/* Finds perspective transformation between the object plane and image (view) plane */
+CVAPI(int) cvFindHomography( const CvMat* src_points,
+                             const CvMat* dst_points,
+                             CvMat* homography,
+                             int method CV_DEFAULT(0),
+                             double ransacReprojThreshold CV_DEFAULT(3),
+                             CvMat* mask CV_DEFAULT(0),
+                             int maxIters CV_DEFAULT(2000),
+                             double confidence CV_DEFAULT(0.995));
+
+/* Computes RQ decomposition for 3x3 matrices */
+CVAPI(void) cvRQDecomp3x3( const CvMat *matrixM, CvMat *matrixR, CvMat *matrixQ,
+                           CvMat *matrixQx CV_DEFAULT(NULL),
+                           CvMat *matrixQy CV_DEFAULT(NULL),
+                           CvMat *matrixQz CV_DEFAULT(NULL),
+                           CvPoint3D64f *eulerAngles CV_DEFAULT(NULL));
+
+/* Computes projection matrix decomposition */
+CVAPI(void) cvDecomposeProjectionMatrix( const CvMat *projMatr, CvMat *calibMatr,
+                                         CvMat *rotMatr, CvMat *posVect,
+                                         CvMat *rotMatrX CV_DEFAULT(NULL),
+                                         CvMat *rotMatrY CV_DEFAULT(NULL),
+                                         CvMat *rotMatrZ CV_DEFAULT(NULL),
+                                         CvPoint3D64f *eulerAngles CV_DEFAULT(NULL));
+
+/* Computes d(AB)/dA and d(AB)/dB */
+CVAPI(void) cvCalcMatMulDeriv( const CvMat* A, const CvMat* B, CvMat* dABdA, CvMat* dABdB );
+
+/* Computes r3 = rodrigues(rodrigues(r2)*rodrigues(r1)),
+   t3 = rodrigues(r2)*t1 + t2 and the respective derivatives */
+CVAPI(void) cvComposeRT( const CvMat* _rvec1, const CvMat* _tvec1,
+                         const CvMat* _rvec2, const CvMat* _tvec2,
+                         CvMat* _rvec3, CvMat* _tvec3,
+                         CvMat* dr3dr1 CV_DEFAULT(0), CvMat* dr3dt1 CV_DEFAULT(0),
+                         CvMat* dr3dr2 CV_DEFAULT(0), CvMat* dr3dt2 CV_DEFAULT(0),
+                         CvMat* dt3dr1 CV_DEFAULT(0), CvMat* dt3dt1 CV_DEFAULT(0),
+                         CvMat* dt3dr2 CV_DEFAULT(0), CvMat* dt3dt2 CV_DEFAULT(0) );
+
+/* Projects object points to the view plane using
+   the specified extrinsic and intrinsic camera parameters */
+CVAPI(void) cvProjectPoints2( const CvMat* object_points, const CvMat* rotation_vector,
+                              const CvMat* translation_vector, const CvMat* camera_matrix,
+                              const CvMat* distortion_coeffs, CvMat* image_points,
+                              CvMat* dpdrot CV_DEFAULT(NULL), CvMat* dpdt CV_DEFAULT(NULL),
+                              CvMat* dpdf CV_DEFAULT(NULL), CvMat* dpdc CV_DEFAULT(NULL),
+                              CvMat* dpddist CV_DEFAULT(NULL),
+                              double aspect_ratio CV_DEFAULT(0));
+
+/* Finds extrinsic camera parameters from
+   a few known corresponding point pairs and intrinsic parameters */
+CVAPI(void) cvFindExtrinsicCameraParams2( const CvMat* object_points,
+                                          const CvMat* image_points,
+                                          const CvMat* camera_matrix,
+                                          const CvMat* distortion_coeffs,
+                                          CvMat* rotation_vector,
+                                          CvMat* translation_vector,
+                                          int use_extrinsic_guess CV_DEFAULT(0) );
+
+/* Computes initial estimate of the intrinsic camera parameters
+   in case of planar calibration target (e.g. chessboard) */
+CVAPI(void) cvInitIntrinsicParams2D( const CvMat* object_points,
+                                     const CvMat* image_points,
+                                     const CvMat* npoints, CvSize image_size,
+                                     CvMat* camera_matrix,
+                                     double aspect_ratio CV_DEFAULT(1.) );
+
+#define CV_CALIB_CB_ADAPTIVE_THRESH  1
+#define CV_CALIB_CB_NORMALIZE_IMAGE  2
+#define CV_CALIB_CB_FILTER_QUADS     4
+#define CV_CALIB_CB_FAST_CHECK       8
+
+// Performs a fast check if a chessboard is in the input image. This is a workaround to
+// a problem of cvFindChessboardCorners being slow on images with no chessboard
+// - src: input image
+// - size: chessboard size
+// Returns 1 if a chessboard can be in this image and findChessboardCorners should be called,
+// 0 if there is no chessboard, -1 in case of error
+CVAPI(int) cvCheckChessboard(IplImage* src, CvSize size);
+
+    /* Detects corners on a chessboard calibration pattern */
+CVAPI(int) cvFindChessboardCorners( const void* image, CvSize pattern_size,
+                                    CvPoint2D32f* corners,
+                                    int* corner_count CV_DEFAULT(NULL),
+                                    int flags CV_DEFAULT(CV_CALIB_CB_ADAPTIVE_THRESH+CV_CALIB_CB_NORMALIZE_IMAGE) );
+
+/* Draws individual chessboard corners or the whole chessboard detected */
+CVAPI(void) cvDrawChessboardCorners( CvArr* image, CvSize pattern_size,
+                                     CvPoint2D32f* corners,
+                                     int count, int pattern_was_found );
+
+#define CV_CALIB_USE_INTRINSIC_GUESS  1
+#define CV_CALIB_FIX_ASPECT_RATIO     2
+#define CV_CALIB_FIX_PRINCIPAL_POINT  4
+#define CV_CALIB_ZERO_TANGENT_DIST    8
+#define CV_CALIB_FIX_FOCAL_LENGTH 16
+#define CV_CALIB_FIX_K1  32
+#define CV_CALIB_FIX_K2  64
+#define CV_CALIB_FIX_K3  128
+#define CV_CALIB_FIX_K4  2048
+#define CV_CALIB_FIX_K5  4096
+#define CV_CALIB_FIX_K6  8192
+#define CV_CALIB_RATIONAL_MODEL 16384
+#define CV_CALIB_THIN_PRISM_MODEL 32768
+#define CV_CALIB_FIX_S1_S2_S3_S4  65536
+#define CV_CALIB_TILTED_MODEL  262144
+#define CV_CALIB_FIX_TAUX_TAUY  524288
+#define CV_CALIB_FIX_TANGENT_DIST 2097152
+
+#define CV_CALIB_NINTRINSIC 18
+
+/* Finds intrinsic and extrinsic camera parameters
+   from a few views of known calibration pattern */
+CVAPI(double) cvCalibrateCamera2( const CvMat* object_points,
+                                const CvMat* image_points,
+                                const CvMat* point_counts,
+                                CvSize image_size,
+                                CvMat* camera_matrix,
+                                CvMat* distortion_coeffs,
+                                CvMat* rotation_vectors CV_DEFAULT(NULL),
+                                CvMat* translation_vectors CV_DEFAULT(NULL),
+                                int flags CV_DEFAULT(0),
+                                CvTermCriteria term_crit CV_DEFAULT(cvTermCriteria(
+                                    CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,DBL_EPSILON)) );
+
+/* Computes various useful characteristics of the camera from the data computed by
+   cvCalibrateCamera2 */
+CVAPI(void) cvCalibrationMatrixValues( const CvMat *camera_matrix,
+                                CvSize image_size,
+                                double aperture_width CV_DEFAULT(0),
+                                double aperture_height CV_DEFAULT(0),
+                                double *fovx CV_DEFAULT(NULL),
+                                double *fovy CV_DEFAULT(NULL),
+                                double *focal_length CV_DEFAULT(NULL),
+                                CvPoint2D64f *principal_point CV_DEFAULT(NULL),
+                                double *pixel_aspect_ratio CV_DEFAULT(NULL));
+
+#define CV_CALIB_FIX_INTRINSIC  256
+#define CV_CALIB_SAME_FOCAL_LENGTH 512
+
+/* Computes the transformation from one camera coordinate system to another one
+   from a few correspondent views of the same calibration target. Optionally, calibrates
+   both cameras */
+CVAPI(double) cvStereoCalibrate( const CvMat* object_points, const CvMat* image_points1,
+                               const CvMat* image_points2, const CvMat* npoints,
+                               CvMat* camera_matrix1, CvMat* dist_coeffs1,
+                               CvMat* camera_matrix2, CvMat* dist_coeffs2,
+                               CvSize image_size, CvMat* R, CvMat* T,
+                               CvMat* E CV_DEFAULT(0), CvMat* F CV_DEFAULT(0),
+                               int flags CV_DEFAULT(CV_CALIB_FIX_INTRINSIC),
+                               CvTermCriteria term_crit CV_DEFAULT(cvTermCriteria(
+                                   CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,1e-6)) );
+
+#define CV_CALIB_ZERO_DISPARITY 1024
+
+/* Computes 3D rotations (+ optional shift) for each camera coordinate system to make both
+   views parallel (=> to make all the epipolar lines horizontal or vertical) */
+CVAPI(void) cvStereoRectify( const CvMat* camera_matrix1, const CvMat* camera_matrix2,
+                             const CvMat* dist_coeffs1, const CvMat* dist_coeffs2,
+                             CvSize image_size, const CvMat* R, const CvMat* T,
+                             CvMat* R1, CvMat* R2, CvMat* P1, CvMat* P2,
+                             CvMat* Q CV_DEFAULT(0),
+                             int flags CV_DEFAULT(CV_CALIB_ZERO_DISPARITY),
+                             double alpha CV_DEFAULT(-1),
+                             CvSize new_image_size CV_DEFAULT(cvSize(0,0)),
+                             CvRect* valid_pix_ROI1 CV_DEFAULT(0),
+                             CvRect* valid_pix_ROI2 CV_DEFAULT(0));
+
+/* Computes rectification transformations for uncalibrated pair of images using a set
+   of point correspondences */
+CVAPI(int) cvStereoRectifyUncalibrated( const CvMat* points1, const CvMat* points2,
+                                        const CvMat* F, CvSize img_size,
+                                        CvMat* H1, CvMat* H2,
+                                        double threshold CV_DEFAULT(5));
+
+
+
+/* stereo correspondence parameters and functions */
+
+#define CV_STEREO_BM_NORMALIZED_RESPONSE  0
+#define CV_STEREO_BM_XSOBEL               1
+
+/* Block matching algorithm structure */
+typedef struct CvStereoBMState
+{
+    // pre-filtering (normalization of input images)
+    int preFilterType; // =CV_STEREO_BM_NORMALIZED_RESPONSE now
+    int preFilterSize; // averaging window size: ~5x5..21x21
+    int preFilterCap; // the output of pre-filtering is clipped by [-preFilterCap,preFilterCap]
+
+    // correspondence using Sum of Absolute Difference (SAD)
+    int SADWindowSize; // ~5x5..21x21
+    int minDisparity;  // minimum disparity (can be negative)
+    int numberOfDisparities; // maximum disparity - minimum disparity (> 0)
+
+    // post-filtering
+    int textureThreshold;  // the disparity is only computed for pixels
+                           // with textured enough neighborhood
+    int uniquenessRatio;   // accept the computed disparity d* only if
+                           // SAD(d) >= SAD(d*)*(1 + uniquenessRatio/100.)
+                           // for any d != d*+/-1 within the search range.
+    int speckleWindowSize; // disparity variation window
+    int speckleRange; // acceptable range of variation in window
+
+    int trySmallerWindows; // if 1, the results may be more accurate,
+                           // at the expense of slower processing
+    CvRect roi1, roi2;
+    int disp12MaxDiff;
+
+    // temporary buffers
+    CvMat* preFilteredImg0;
+    CvMat* preFilteredImg1;
+    CvMat* slidingSumBuf;
+    CvMat* cost;
+    CvMat* disp;
+} CvStereoBMState;
+
+#define CV_STEREO_BM_BASIC 0
+#define CV_STEREO_BM_FISH_EYE 1
+#define CV_STEREO_BM_NARROW 2
+
+CVAPI(CvStereoBMState*) cvCreateStereoBMState(int preset CV_DEFAULT(CV_STEREO_BM_BASIC),
+                                              int numberOfDisparities CV_DEFAULT(0));
+
+CVAPI(void) cvReleaseStereoBMState( CvStereoBMState** state );
+
+CVAPI(void) cvFindStereoCorrespondenceBM( const CvArr* left, const CvArr* right,
+                                          CvArr* disparity, CvStereoBMState* state );
+
+CVAPI(CvRect) cvGetValidDisparityROI( CvRect roi1, CvRect roi2, int minDisparity,
+                                      int numberOfDisparities, int SADWindowSize );
+
+CVAPI(void) cvValidateDisparity( CvArr* disparity, const CvArr* cost,
+                                 int minDisparity, int numberOfDisparities,
+                                 int disp12MaxDiff CV_DEFAULT(1) );
+
+/* Reprojects the computed disparity image to the 3D space using the specified 4x4 matrix */
+CVAPI(void)  cvReprojectImageTo3D( const CvArr* disparityImage,
+                                   CvArr* _3dImage, const CvMat* Q,
+                                   int handleMissingValues CV_DEFAULT(0) );
+
+/** @} calib3d_c */
+
+#ifdef __cplusplus
+} // extern "C"
+
+//////////////////////////////////////////////////////////////////////////////////////////
+class CV_EXPORTS CvLevMarq
+{
+public:
+    CvLevMarq();
+    CvLevMarq( int nparams, int nerrs, CvTermCriteria criteria=
+              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
+              bool completeSymmFlag=false );
+    ~CvLevMarq();
+    void init( int nparams, int nerrs, CvTermCriteria criteria=
+              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
+              bool completeSymmFlag=false );
+    bool update( const CvMat*& param, CvMat*& J, CvMat*& err );
+    bool updateAlt( const CvMat*& param, CvMat*& JtJ, CvMat*& JtErr, double*& errNorm );
+
+    void clear();
+    void step();
+    enum { DONE=0, STARTED=1, CALC_J=2, CHECK_ERR=3 };
+
+    cv::Ptr<CvMat> mask;
+    cv::Ptr<CvMat> prevParam;
+    cv::Ptr<CvMat> param;
+    cv::Ptr<CvMat> J;
+    cv::Ptr<CvMat> err;
+    cv::Ptr<CvMat> JtJ;
+    cv::Ptr<CvMat> JtJN;
+    cv::Ptr<CvMat> JtErr;
+    cv::Ptr<CvMat> JtJV;
+    cv::Ptr<CvMat> JtJW;
+    double prevErrNorm, errNorm;
+    int lambdaLg10;
+    CvTermCriteria criteria;
+    int state;
+    int iters;
+    bool completeSymmFlag;
+    int solveMethod;
+};
+
+#endif
+
+#endif /* OPENCV_CALIB3D_C_H */

3rdPart/include/opencv2/core/affine.hpp

@@ -0,0 +1,678 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_AFFINE3_HPP
+#define OPENCV_CORE_AFFINE3_HPP
+
+#ifdef __cplusplus
+
+#include <opencv2/core.hpp>
+
+namespace cv
+{
+
+//! @addtogroup core
+//! @{
+
+    /** @brief Affine transform
+     *
+     * It represents a 4x4 homogeneous transformation matrix \f$T\f$
+     *
+     *  \f[T =
+     *  \begin{bmatrix}
+     *  R & t\\
+     *  0 & 1\\
+     *  \end{bmatrix}
+     *  \f]
+     *
+     *  where \f$R\f$ is a 3x3 rotation matrix and \f$t\f$ is a 3x1 translation vector.
+     *
+     *  You can specify \f$R\f$ either by a 3x3 rotation matrix or by a 3x1 rotation vector,
+     *  which is converted to a 3x3 rotation matrix by the Rodrigues formula.
+     *
+     *  To construct a matrix \f$T\f$ representing first rotation around the axis \f$r\f$ with rotation
+     *  angle \f$|r|\f$ in radian (right hand rule) and then translation by the vector \f$t\f$, you can use
+     *
+     *  @code
+     *  cv::Vec3f r, t;
+     *  cv::Affine3f T(r, t);
+     *  @endcode
+     *
+     *  If you already have the rotation matrix \f$R\f$, then you can use
+     *
+     *  @code
+     *  cv::Matx33f R;
+     *  cv::Affine3f T(R, t);
+     *  @endcode
+     *
+     *  To extract the rotation matrix \f$R\f$ from \f$T\f$, use
+     *
+     *  @code
+     *  cv::Matx33f R = T.rotation();
+     *  @endcode
+     *
+     *  To extract the translation vector \f$t\f$ from \f$T\f$, use
+     *
+     *  @code
+     *  cv::Vec3f t = T.translation();
+     *  @endcode
+     *
+     *  To extract the rotation vector \f$r\f$ from \f$T\f$, use
+     *
+     *  @code
+     *  cv::Vec3f r = T.rvec();
+     *  @endcode
+     *
+     *  Note that since the mapping from rotation vectors to rotation matrices
+     *  is many to one. The returned rotation vector is not necessarily the one
+     *  you used before to set the matrix.
+     *
+     *  If you have two transformations \f$T = T_1 * T_2\f$, use
+     *
+     *  @code
+     *  cv::Affine3f T, T1, T2;
+     *  T = T2.concatenate(T1);
+     *  @endcode
+     *
+     *  To get the inverse transform of \f$T\f$, use
+     *
+     *  @code
+     *  cv::Affine3f T, T_inv;
+     *  T_inv = T.inv();
+     *  @endcode
+     *
+     */
+    template<typename T>
+    class Affine3
+    {
+    public:
+        typedef T float_type;
+        typedef Matx<float_type, 3, 3> Mat3;
+        typedef Matx<float_type, 4, 4> Mat4;
+        typedef Vec<float_type, 3> Vec3;
+
+       //! Default constructor. It represents a 4x4 identity matrix.
+        Affine3();
+
+        //! Augmented affine matrix
+        Affine3(const Mat4& affine);
+
+        /**
+         *  The resulting 4x4 matrix is
+         *
+         *  \f[
+         *  \begin{bmatrix}
+         *  R & t\\
+         *  0 & 1\\
+         *  \end{bmatrix}
+         *  \f]
+         *
+         * @param R 3x3 rotation matrix.
+         * @param t 3x1 translation vector.
+         */
+        Affine3(const Mat3& R, const Vec3& t = Vec3::all(0));
+
+        /**
+         * Rodrigues vector.
+         *
+         * The last row of the current matrix is set to [0,0,0,1].
+         *
+         * @param rvec 3x1 rotation vector. Its direction indicates the rotation axis and its length
+         *             indicates the rotation angle in radian (using right hand rule).
+         * @param t 3x1 translation vector.
+         */
+        Affine3(const Vec3& rvec, const Vec3& t = Vec3::all(0));
+
+        /**
+         * Combines all constructors above. Supports 4x4, 3x4, 3x3, 1x3, 3x1 sizes of data matrix.
+         *
+         * The last row of the current matrix is set to [0,0,0,1] when data is not 4x4.
+         *
+         * @param data 1-channel matrix.
+         *             when it is 4x4, it is copied to the current matrix and t is not used.
+         *             When it is 3x4, it is copied to the upper part 3x4 of the current matrix and t is not used.
+         *             When it is 3x3, it is copied to the upper left 3x3 part of the current matrix.
+         *             When it is 3x1 or 1x3, it is treated as a rotation vector and the Rodrigues formula is used
+         *                             to compute a 3x3 rotation matrix.
+         * @param t 3x1 translation vector. It is used only when data is neither 4x4 nor 3x4.
+         */
+        explicit Affine3(const Mat& data, const Vec3& t = Vec3::all(0));
+
+        //! From 16-element array
+        explicit Affine3(const float_type* vals);
+
+        //! Create an 4x4 identity transform
+        static Affine3 Identity();
+
+        /**
+         * Rotation matrix.
+         *
+         * Copy the rotation matrix to the upper left 3x3 part of the current matrix.
+         * The remaining elements of the current matrix are not changed.
+         *
+         * @param R 3x3 rotation matrix.
+         *
+         */
+        void rotation(const Mat3& R);
+
+        /**
+         * Rodrigues vector.
+         *
+         * It sets the upper left 3x3 part of the matrix. The remaining part is unaffected.
+         *
+         * @param rvec 3x1 rotation vector. The direction indicates the rotation axis and
+         *             its length indicates the rotation angle in radian (using the right thumb convention).
+         */
+        void rotation(const Vec3& rvec);
+
+        /**
+         * Combines rotation methods above. Supports 3x3, 1x3, 3x1 sizes of data matrix.
+         *
+         * It sets the upper left 3x3 part of the matrix. The remaining part is unaffected.
+         *
+         * @param data 1-channel matrix.
+         *             When it is a 3x3 matrix, it sets the upper left 3x3 part of the current matrix.
+         *             When it is a 1x3 or 3x1 matrix, it is used as a rotation vector. The Rodrigues formula
+         *             is used to compute the rotation matrix and sets the upper left 3x3 part of the current matrix.
+         */
+        void rotation(const Mat& data);
+
+        /**
+         * Copy the 3x3 matrix L to the upper left part of the current matrix
+         *
+         * It sets the upper left 3x3 part of the matrix. The remaining part is unaffected.
+         *
+         * @param L 3x3 matrix.
+         */
+        void linear(const Mat3& L);
+
+        /**
+         * Copy t to the first three elements of the last column of the current matrix
+         *
+         * It sets the upper right 3x1 part of the matrix. The remaining part is unaffected.
+         *
+         * @param t 3x1 translation vector.
+         */
+        void translation(const Vec3& t);
+
+        //! @return the upper left 3x3 part
+        Mat3 rotation() const;
+
+        //! @return the upper left 3x3 part
+        Mat3 linear() const;
+
+        //! @return the upper right 3x1 part
+        Vec3 translation() const;
+
+        //! Rodrigues vector.
+        //! @return a vector representing the upper left 3x3 rotation matrix of the current matrix.
+        //! @warning  Since the mapping between rotation vectors and rotation matrices is many to one,
+        //!           this function returns only one rotation vector that represents the current rotation matrix,
+        //!           which is not necessarily the same one set by `rotation(const Vec3& rvec)`.
+        Vec3 rvec() const;
+
+        //! @return the inverse of the current matrix.
+        Affine3 inv(int method = cv::DECOMP_SVD) const;
+
+        //! a.rotate(R) is equivalent to Affine(R, 0) * a;
+        Affine3 rotate(const Mat3& R) const;
+
+        //! a.rotate(rvec) is equivalent to Affine(rvec, 0) * a;
+        Affine3 rotate(const Vec3& rvec) const;
+
+        //! a.translate(t) is equivalent to Affine(E, t) * a, where E is an identity matrix
+        Affine3 translate(const Vec3& t) const;
+
+        //! a.concatenate(affine) is equivalent to affine * a;
+        Affine3 concatenate(const Affine3& affine) const;
+
+        template <typename Y> operator Affine3<Y>() const;
+
+        template <typename Y> Affine3<Y> cast() const;
+
+        Mat4 matrix;
+
+#if defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H
+        Affine3(const Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>& affine);
+        Affine3(const Eigen::Transform<T, 3, Eigen::Affine>& affine);
+        operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>() const;
+        operator Eigen::Transform<T, 3, Eigen::Affine>() const;
+#endif
+    };
+
+    template<typename T> static
+    Affine3<T> operator*(const Affine3<T>& affine1, const Affine3<T>& affine2);
+
+    //! V is a 3-element vector with member fields x, y and z
+    template<typename T, typename V> static
+    V operator*(const Affine3<T>& affine, const V& vector);
+
+    typedef Affine3<float> Affine3f;
+    typedef Affine3<double> Affine3d;
+
+    static Vec3f operator*(const Affine3f& affine, const Vec3f& vector);
+    static Vec3d operator*(const Affine3d& affine, const Vec3d& vector);
+
+    template<typename _Tp> class DataType< Affine3<_Tp> >
+    {
+    public:
+        typedef Affine3<_Tp>                               value_type;
+        typedef Affine3<typename DataType<_Tp>::work_type> work_type;
+        typedef _Tp                                        channel_type;
+
+        enum { generic_type = 0,
+               channels     = 16,
+               fmt          = traits::SafeFmt<channel_type>::fmt + ((channels - 1) << 8)
+#ifdef OPENCV_TRAITS_ENABLE_DEPRECATED
+               ,depth        = DataType<channel_type>::depth
+               ,type         = CV_MAKETYPE(depth, channels)
+#endif
+             };
+
+        typedef Vec<channel_type, channels> vec_type;
+    };
+
+    namespace traits {
+    template<typename _Tp>
+    struct Depth< Affine3<_Tp> > { enum { value = Depth<_Tp>::value }; };
+    template<typename _Tp>
+    struct Type< Affine3<_Tp> > { enum { value = CV_MAKETYPE(Depth<_Tp>::value, 16) }; };
+    } // namespace
+
+//! @} core
+
+}
+
+//! @cond IGNORED
+
+///////////////////////////////////////////////////////////////////////////////////
+// Implementation
+
+template<typename T> inline
+cv::Affine3<T>::Affine3()
+    : matrix(Mat4::eye())
+{}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Mat4& affine)
+    : matrix(affine)
+{}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Mat3& R, const Vec3& t)
+{
+    rotation(R);
+    translation(t);
+    matrix.val[12] = matrix.val[13] = matrix.val[14] = 0;
+    matrix.val[15] = 1;
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Vec3& _rvec, const Vec3& t)
+{
+    rotation(_rvec);
+    translation(t);
+    matrix.val[12] = matrix.val[13] = matrix.val[14] = 0;
+    matrix.val[15] = 1;
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const cv::Mat& data, const Vec3& t)
+{
+    CV_Assert(data.type() == cv::traits::Type<T>::value);
+    CV_Assert(data.channels() == 1);
+
+    if (data.cols == 4 && data.rows == 4)
+    {
+        data.copyTo(matrix);
+        return;
+    }
+    else if (data.cols == 4 && data.rows == 3)
+    {
+        rotation(data(Rect(0, 0, 3, 3)));
+        translation(data(Rect(3, 0, 1, 3)));
+    }
+    else
+    {
+        rotation(data);
+        translation(t);
+    }
+
+    matrix.val[12] = matrix.val[13] = matrix.val[14] = 0;
+    matrix.val[15] = 1;
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const float_type* vals) : matrix(vals)
+{}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::Identity()
+{
+    return Affine3<T>(cv::Affine3<T>::Mat4::eye());
+}
+
+template<typename T> inline
+void cv::Affine3<T>::rotation(const Mat3& R)
+{
+    linear(R);
+}
+
+template<typename T> inline
+void cv::Affine3<T>::rotation(const Vec3& _rvec)
+{
+    double theta = norm(_rvec);
+
+    if (theta < DBL_EPSILON)
+        rotation(Mat3::eye());
+    else
+    {
+        double c = std::cos(theta);
+        double s = std::sin(theta);
+        double c1 = 1. - c;
+        double itheta = (theta != 0) ? 1./theta : 0.;
+
+        Point3_<T> r = _rvec*itheta;
+
+        Mat3 rrt( r.x*r.x, r.x*r.y, r.x*r.z, r.x*r.y, r.y*r.y, r.y*r.z, r.x*r.z, r.y*r.z, r.z*r.z );
+        Mat3 r_x( 0, -r.z, r.y, r.z, 0, -r.x, -r.y, r.x, 0 );
+
+        // R = cos(theta)*I + (1 - cos(theta))*r*rT + sin(theta)*[r_x]
+        // where [r_x] is [0 -rz ry; rz 0 -rx; -ry rx 0]
+        Mat3 R = c*Mat3::eye() + c1*rrt + s*r_x;
+
+        rotation(R);
+    }
+}
+
+//Combines rotation methods above. Supports 3x3, 1x3, 3x1 sizes of data matrix;
+template<typename T> inline
+void cv::Affine3<T>::rotation(const cv::Mat& data)
+{
+    CV_Assert(data.type() == cv::traits::Type<T>::value);
+    CV_Assert(data.channels() == 1);
+
+    if (data.cols == 3 && data.rows == 3)
+    {
+        Mat3 R;
+        data.copyTo(R);
+        rotation(R);
+    }
+    else if ((data.cols == 3 && data.rows == 1) || (data.cols == 1 && data.rows == 3))
+    {
+        Vec3 _rvec;
+        data.reshape(1, 3).copyTo(_rvec);
+        rotation(_rvec);
+    }
+    else
+        CV_Error(Error::StsError, "Input matrix can only be 3x3, 1x3 or 3x1");
+}
+
+template<typename T> inline
+void cv::Affine3<T>::linear(const Mat3& L)
+{
+    matrix.val[0] = L.val[0]; matrix.val[1] = L.val[1];  matrix.val[ 2] = L.val[2];
+    matrix.val[4] = L.val[3]; matrix.val[5] = L.val[4];  matrix.val[ 6] = L.val[5];
+    matrix.val[8] = L.val[6]; matrix.val[9] = L.val[7];  matrix.val[10] = L.val[8];
+}
+
+template<typename T> inline
+void cv::Affine3<T>::translation(const Vec3& t)
+{
+    matrix.val[3] = t[0]; matrix.val[7] = t[1]; matrix.val[11] = t[2];
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Mat3 cv::Affine3<T>::rotation() const
+{
+    return linear();
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Mat3 cv::Affine3<T>::linear() const
+{
+    typename cv::Affine3<T>::Mat3 R;
+    R.val[0] = matrix.val[0];  R.val[1] = matrix.val[1];  R.val[2] = matrix.val[ 2];
+    R.val[3] = matrix.val[4];  R.val[4] = matrix.val[5];  R.val[5] = matrix.val[ 6];
+    R.val[6] = matrix.val[8];  R.val[7] = matrix.val[9];  R.val[8] = matrix.val[10];
+    return R;
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Vec3 cv::Affine3<T>::translation() const
+{
+    return Vec3(matrix.val[3], matrix.val[7], matrix.val[11]);
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Vec3 cv::Affine3<T>::rvec() const
+{
+    cv::Vec3d w;
+    cv::Matx33d u, vt, R = rotation();
+    cv::SVD::compute(R, w, u, vt, cv::SVD::FULL_UV + cv::SVD::MODIFY_A);
+    R = u * vt;
+
+    double rx = R.val[7] - R.val[5];
+    double ry = R.val[2] - R.val[6];
+    double rz = R.val[3] - R.val[1];
+
+    double s = std::sqrt((rx*rx + ry*ry + rz*rz)*0.25);
+    double c = (R.val[0] + R.val[4] + R.val[8] - 1) * 0.5;
+    c = c > 1.0 ? 1.0 : c < -1.0 ? -1.0 : c;
+    double theta = acos(c);
+
+    if( s < 1e-5 )
+    {
+        if( c > 0 )
+            rx = ry = rz = 0;
+        else
+        {
+            double t;
+            t = (R.val[0] + 1) * 0.5;
+            rx = std::sqrt(std::max(t, 0.0));
+            t = (R.val[4] + 1) * 0.5;
+            ry = std::sqrt(std::max(t, 0.0)) * (R.val[1] < 0 ? -1.0 : 1.0);
+            t = (R.val[8] + 1) * 0.5;
+            rz = std::sqrt(std::max(t, 0.0)) * (R.val[2] < 0 ? -1.0 : 1.0);
+
+            if( fabs(rx) < fabs(ry) && fabs(rx) < fabs(rz) && (R.val[5] > 0) != (ry*rz > 0) )
+                rz = -rz;
+            theta /= std::sqrt(rx*rx + ry*ry + rz*rz);
+            rx *= theta;
+            ry *= theta;
+            rz *= theta;
+        }
+    }
+    else
+    {
+        double vth = 1/(2*s);
+        vth *= theta;
+        rx *= vth; ry *= vth; rz *= vth;
+    }
+
+    return cv::Vec3d(rx, ry, rz);
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::inv(int method) const
+{
+    return matrix.inv(method);
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::rotate(const Mat3& R) const
+{
+    Mat3 Lc = linear();
+    Vec3 tc = translation();
+    Mat4 result;
+    result.val[12] = result.val[13] = result.val[14] = 0;
+    result.val[15] = 1;
+
+    for(int j = 0; j < 3; ++j)
+    {
+        for(int i = 0; i < 3; ++i)
+        {
+            float_type value = 0;
+            for(int k = 0; k < 3; ++k)
+                value += R(j, k) * Lc(k, i);
+            result(j, i) = value;
+        }
+
+        result(j, 3) = R.row(j).dot(tc.t());
+    }
+    return result;
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::rotate(const Vec3& _rvec) const
+{
+    return rotate(Affine3f(_rvec).rotation());
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::translate(const Vec3& t) const
+{
+    Mat4 m = matrix;
+    m.val[ 3] += t[0];
+    m.val[ 7] += t[1];
+    m.val[11] += t[2];
+    return m;
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::concatenate(const Affine3<T>& affine) const
+{
+    return (*this).rotate(affine.rotation()).translate(affine.translation());
+}
+
+template<typename T> template <typename Y> inline
+cv::Affine3<T>::operator Affine3<Y>() const
+{
+    return Affine3<Y>(matrix);
+}
+
+template<typename T> template <typename Y> inline
+cv::Affine3<Y> cv::Affine3<T>::cast() const
+{
+    return Affine3<Y>(matrix);
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::operator*(const cv::Affine3<T>& affine1, const cv::Affine3<T>& affine2)
+{
+    return affine2.concatenate(affine1);
+}
+
+template<typename T, typename V> inline
+V cv::operator*(const cv::Affine3<T>& affine, const V& v)
+{
+    const typename Affine3<T>::Mat4& m = affine.matrix;
+
+    V r;
+    r.x = m.val[0] * v.x + m.val[1] * v.y + m.val[ 2] * v.z + m.val[ 3];
+    r.y = m.val[4] * v.x + m.val[5] * v.y + m.val[ 6] * v.z + m.val[ 7];
+    r.z = m.val[8] * v.x + m.val[9] * v.y + m.val[10] * v.z + m.val[11];
+    return r;
+}
+
+static inline
+cv::Vec3f cv::operator*(const cv::Affine3f& affine, const cv::Vec3f& v)
+{
+    const cv::Matx44f& m = affine.matrix;
+    cv::Vec3f r;
+    r.val[0] = m.val[0] * v[0] + m.val[1] * v[1] + m.val[ 2] * v[2] + m.val[ 3];
+    r.val[1] = m.val[4] * v[0] + m.val[5] * v[1] + m.val[ 6] * v[2] + m.val[ 7];
+    r.val[2] = m.val[8] * v[0] + m.val[9] * v[1] + m.val[10] * v[2] + m.val[11];
+    return r;
+}
+
+static inline
+cv::Vec3d cv::operator*(const cv::Affine3d& affine, const cv::Vec3d& v)
+{
+    const cv::Matx44d& m = affine.matrix;
+    cv::Vec3d r;
+    r.val[0] = m.val[0] * v[0] + m.val[1] * v[1] + m.val[ 2] * v[2] + m.val[ 3];
+    r.val[1] = m.val[4] * v[0] + m.val[5] * v[1] + m.val[ 6] * v[2] + m.val[ 7];
+    r.val[2] = m.val[8] * v[0] + m.val[9] * v[1] + m.val[10] * v[2] + m.val[11];
+    return r;
+}
+
+
+
+#if defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>& affine)
+{
+    cv::Mat(4, 4, cv::traits::Type<T>::value, affine.matrix().data()).copyTo(matrix);
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine>& affine)
+{
+    Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)> a = affine;
+    cv::Mat(4, 4, cv::traits::Type<T>::value, a.matrix().data()).copyTo(matrix);
+}
+
+template<typename T> inline
+cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>() const
+{
+    Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)> r;
+    cv::Mat hdr(4, 4, cv::traits::Type<T>::value, r.matrix().data());
+    cv::Mat(matrix, false).copyTo(hdr);
+    return r;
+}
+
+template<typename T> inline
+cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine>() const
+{
+    return this->operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>();
+}
+
+#endif /* defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H */
+
+//! @endcond
+
+#endif /* __cplusplus */
+
+#endif /* OPENCV_CORE_AFFINE3_HPP */

3rdPart/include/opencv2/core/async.hpp

@@ -0,0 +1,105 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_ASYNC_HPP
+#define OPENCV_CORE_ASYNC_HPP
+
+#include <opencv2/core/mat.hpp>
+
+#ifdef CV_CXX11
+//#include <future>
+#include <chrono>
+#endif
+
+namespace cv {
+
+/** @addtogroup core_async
+
+@{
+*/
+
+
+/** @brief Returns result of asynchronous operations
+
+Object has attached asynchronous state.
+Assignment operator doesn't clone asynchronous state (it is shared between all instances).
+
+Result can be fetched via get() method only once.
+
+*/
+class CV_EXPORTS_W AsyncArray
+{
+public:
+    ~AsyncArray() CV_NOEXCEPT;
+    CV_WRAP AsyncArray() CV_NOEXCEPT;
+    AsyncArray(const AsyncArray& o) CV_NOEXCEPT;
+    AsyncArray& operator=(const AsyncArray& o) CV_NOEXCEPT;
+    CV_WRAP void release() CV_NOEXCEPT;
+
+    /** Fetch the result.
+    @param[out] dst destination array
+
+    Waits for result until container has valid result.
+    Throws exception if exception was stored as a result.
+
+    Throws exception on invalid container state.
+
+    @note Result or stored exception can be fetched only once.
+    */
+    CV_WRAP void get(OutputArray dst) const;
+
+    /** Retrieving the result with timeout
+    @param[out] dst destination array
+    @param[in] timeoutNs timeout in nanoseconds, -1 for infinite wait
+
+    @returns true if result is ready, false if the timeout has expired
+
+    @note Result or stored exception can be fetched only once.
+    */
+    bool get(OutputArray dst, int64 timeoutNs) const;
+
+    CV_WRAP inline
+    bool get(OutputArray dst, double timeoutNs) const { return get(dst, (int64)timeoutNs); }
+
+    bool wait_for(int64 timeoutNs) const;
+
+    CV_WRAP inline
+    bool wait_for(double timeoutNs) const { return wait_for((int64)timeoutNs); }
+
+    CV_WRAP bool valid() const CV_NOEXCEPT;
+
+#ifdef CV_CXX11
+    inline AsyncArray(AsyncArray&& o) { p = o.p; o.p = NULL; }
+    inline AsyncArray& operator=(AsyncArray&& o) CV_NOEXCEPT { std::swap(p, o.p); return *this; }
+
+    template<typename _Rep, typename _Period>
+    inline bool get(OutputArray dst, const std::chrono::duration<_Rep, _Period>& timeout)
+    {
+        return get(dst, (int64)(std::chrono::nanoseconds(timeout).count()));
+    }
+
+    template<typename _Rep, typename _Period>
+    inline bool wait_for(const std::chrono::duration<_Rep, _Period>& timeout)
+    {
+        return wait_for((int64)(std::chrono::nanoseconds(timeout).count()));
+    }
+
+#if 0
+    std::future<Mat> getFutureMat() const;
+    std::future<UMat> getFutureUMat() const;
+#endif
+#endif
+
+
+    // PImpl
+    struct Impl; friend struct Impl;
+    inline void* _getImpl() const CV_NOEXCEPT { return p; }
+protected:
+    Impl* p;
+};
+
+
+//! @}
+} // namespace
+#endif // OPENCV_CORE_ASYNC_HPP

3rdPart/include/opencv2/core/base.hpp

@@ -0,0 +1,722 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2014, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_BASE_HPP
+#define OPENCV_CORE_BASE_HPP
+
+#ifndef __cplusplus
+#  error base.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/opencv_modules.hpp"
+
+#include <climits>
+#include <algorithm>
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/cvstd.hpp"
+
+namespace cv
+{
+
+//! @addtogroup core_utils
+//! @{
+
+namespace Error {
+//! error codes
+enum Code {
+    StsOk=                       0,  //!< everything is ok
+    StsBackTrace=               -1,  //!< pseudo error for back trace
+    StsError=                   -2,  //!< unknown /unspecified error
+    StsInternal=                -3,  //!< internal error (bad state)
+    StsNoMem=                   -4,  //!< insufficient memory
+    StsBadArg=                  -5,  //!< function arg/param is bad
+    StsBadFunc=                 -6,  //!< unsupported function
+    StsNoConv=                  -7,  //!< iteration didn't converge
+    StsAutoTrace=               -8,  //!< tracing
+    HeaderIsNull=               -9,  //!< image header is NULL
+    BadImageSize=              -10,  //!< image size is invalid
+    BadOffset=                 -11,  //!< offset is invalid
+    BadDataPtr=                -12,  //!<
+    BadStep=                   -13,  //!< image step is wrong, this may happen for a non-continuous matrix.
+    BadModelOrChSeq=           -14,  //!<
+    BadNumChannels=            -15,  //!< bad number of channels, for example, some functions accept only single channel matrices.
+    BadNumChannel1U=           -16,  //!<
+    BadDepth=                  -17,  //!< input image depth is not supported by the function
+    BadAlphaChannel=           -18,  //!<
+    BadOrder=                  -19,  //!< number of dimensions is out of range
+    BadOrigin=                 -20,  //!< incorrect input origin
+    BadAlign=                  -21,  //!< incorrect input align
+    BadCallBack=               -22,  //!<
+    BadTileSize=               -23,  //!<
+    BadCOI=                    -24,  //!< input COI is not supported
+    BadROISize=                -25,  //!< incorrect input roi
+    MaskIsTiled=               -26,  //!<
+    StsNullPtr=                -27,  //!< null pointer
+    StsVecLengthErr=           -28,  //!< incorrect vector length
+    StsFilterStructContentErr= -29,  //!< incorrect filter structure content
+    StsKernelStructContentErr= -30,  //!< incorrect transform kernel content
+    StsFilterOffsetErr=        -31,  //!< incorrect filter offset value
+    StsBadSize=                -201, //!< the input/output structure size is incorrect
+    StsDivByZero=              -202, //!< division by zero
+    StsInplaceNotSupported=    -203, //!< in-place operation is not supported
+    StsObjectNotFound=         -204, //!< request can't be completed
+    StsUnmatchedFormats=       -205, //!< formats of input/output arrays differ
+    StsBadFlag=                -206, //!< flag is wrong or not supported
+    StsBadPoint=               -207, //!< bad CvPoint
+    StsBadMask=                -208, //!< bad format of mask (neither 8uC1 nor 8sC1)
+    StsUnmatchedSizes=         -209, //!< sizes of input/output structures do not match
+    StsUnsupportedFormat=      -210, //!< the data format/type is not supported by the function
+    StsOutOfRange=             -211, //!< some of parameters are out of range
+    StsParseError=             -212, //!< invalid syntax/structure of the parsed file
+    StsNotImplemented=         -213, //!< the requested function/feature is not implemented
+    StsBadMemBlock=            -214, //!< an allocated block has been corrupted
+    StsAssert=                 -215, //!< assertion failed
+    GpuNotSupported=           -216, //!< no CUDA support
+    GpuApiCallError=           -217, //!< GPU API call error
+    OpenGlNotSupported=        -218, //!< no OpenGL support
+    OpenGlApiCallError=        -219, //!< OpenGL API call error
+    OpenCLApiCallError=        -220, //!< OpenCL API call error
+    OpenCLDoubleNotSupported=  -221,
+    OpenCLInitError=           -222, //!< OpenCL initialization error
+    OpenCLNoAMDBlasFft=        -223
+};
+} //Error
+
+//! @} core_utils
+
+//! @addtogroup core_array
+//! @{
+
+//! matrix decomposition types
+enum DecompTypes {
+    /** Gaussian elimination with the optimal pivot element chosen. */
+    DECOMP_LU       = 0,
+    /** singular value decomposition (SVD) method; the system can be over-defined and/or the matrix
+    src1 can be singular */
+    DECOMP_SVD      = 1,
+    /** eigenvalue decomposition; the matrix src1 must be symmetrical */
+    DECOMP_EIG      = 2,
+    /** Cholesky \f$LL^T\f$ factorization; the matrix src1 must be symmetrical and positively
+    defined */
+    DECOMP_CHOLESKY = 3,
+    /** QR factorization; the system can be over-defined and/or the matrix src1 can be singular */
+    DECOMP_QR       = 4,
+    /** while all the previous flags are mutually exclusive, this flag can be used together with
+    any of the previous; it means that the normal equations
+    \f$\texttt{src1}^T\cdot\texttt{src1}\cdot\texttt{dst}=\texttt{src1}^T\texttt{src2}\f$ are
+    solved instead of the original system
+    \f$\texttt{src1}\cdot\texttt{dst}=\texttt{src2}\f$ */
+    DECOMP_NORMAL   = 16
+};
+
+/** norm types
+
+src1 and src2 denote input arrays.
+*/
+
+enum NormTypes {
+                /**
+                \f[
+                norm =  \forkthree
+                {\|\texttt{src1}\|_{L_{\infty}} =  \max _I | \texttt{src1} (I)|}{if  \(\texttt{normType} = \texttt{NORM_INF}\) }
+                {\|\texttt{src1}-\texttt{src2}\|_{L_{\infty}} =  \max _I | \texttt{src1} (I) -  \texttt{src2} (I)|}{if  \(\texttt{normType} = \texttt{NORM_INF}\) }
+                {\frac{\|\texttt{src1}-\texttt{src2}\|_{L_{\infty}}    }{\|\texttt{src2}\|_{L_{\infty}} }}{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_INF}\) }
+                \f]
+                */
+                NORM_INF       = 1,
+                /**
+                \f[
+                norm =  \forkthree
+                {\| \texttt{src1} \| _{L_1} =  \sum _I | \texttt{src1} (I)|}{if  \(\texttt{normType} = \texttt{NORM_L1}\)}
+                { \| \texttt{src1} - \texttt{src2} \| _{L_1} =  \sum _I | \texttt{src1} (I) -  \texttt{src2} (I)|}{if  \(\texttt{normType} = \texttt{NORM_L1}\) }
+                { \frac{\|\texttt{src1}-\texttt{src2}\|_{L_1} }{\|\texttt{src2}\|_{L_1}} }{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_L1}\) }
+                \f]*/
+                 NORM_L1        = 2,
+                 /**
+                 \f[
+                 norm =  \forkthree
+                 { \| \texttt{src1} \| _{L_2} =  \sqrt{\sum_I \texttt{src1}(I)^2} }{if  \(\texttt{normType} = \texttt{NORM_L2}\) }
+                 { \| \texttt{src1} - \texttt{src2} \| _{L_2} =  \sqrt{\sum_I (\texttt{src1}(I) - \texttt{src2}(I))^2} }{if  \(\texttt{normType} = \texttt{NORM_L2}\) }
+                 { \frac{\|\texttt{src1}-\texttt{src2}\|_{L_2} }{\|\texttt{src2}\|_{L_2}} }{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_L2}\) }
+                 \f]
+                 */
+                 NORM_L2        = 4,
+                 /**
+                 \f[
+                 norm =  \forkthree
+                 { \| \texttt{src1} \| _{L_2} ^{2} = \sum_I \texttt{src1}(I)^2} {if  \(\texttt{normType} = \texttt{NORM_L2SQR}\)}
+                 { \| \texttt{src1} - \texttt{src2} \| _{L_2} ^{2} =  \sum_I (\texttt{src1}(I) - \texttt{src2}(I))^2 }{if  \(\texttt{normType} = \texttt{NORM_L2SQR}\) }
+                 { \left(\frac{\|\texttt{src1}-\texttt{src2}\|_{L_2} }{\|\texttt{src2}\|_{L_2}}\right)^2 }{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_L2SQR}\) }
+                 \f]
+                 */
+                 NORM_L2SQR     = 5,
+                 /**
+                 In the case of one input array, calculates the Hamming distance of the array from zero,
+                 In the case of two input arrays, calculates the Hamming distance between the arrays.
+                 */
+                 NORM_HAMMING   = 6,
+                 /**
+                 Similar to NORM_HAMMING, but in the calculation, each two bits of the input sequence will
+                 be added and treated as a single bit to be used in the same calculation as NORM_HAMMING.
+                 */
+                 NORM_HAMMING2  = 7,
+                 NORM_TYPE_MASK = 7, //!< bit-mask which can be used to separate norm type from norm flags
+                 NORM_RELATIVE  = 8, //!< flag
+                 NORM_MINMAX    = 32 //!< flag
+               };
+
+//! comparison types
+enum CmpTypes { CMP_EQ = 0, //!< src1 is equal to src2.
+                CMP_GT = 1, //!< src1 is greater than src2.
+                CMP_GE = 2, //!< src1 is greater than or equal to src2.
+                CMP_LT = 3, //!< src1 is less than src2.
+                CMP_LE = 4, //!< src1 is less than or equal to src2.
+                CMP_NE = 5  //!< src1 is unequal to src2.
+              };
+
+//! generalized matrix multiplication flags
+enum GemmFlags { GEMM_1_T = 1, //!< transposes src1
+                 GEMM_2_T = 2, //!< transposes src2
+                 GEMM_3_T = 4 //!< transposes src3
+               };
+
+enum DftFlags {
+    /** performs an inverse 1D or 2D transform instead of the default forward
+        transform. */
+    DFT_INVERSE        = 1,
+    /** scales the result: divide it by the number of array elements. Normally, it is
+        combined with DFT_INVERSE. */
+    DFT_SCALE          = 2,
+    /** performs a forward or inverse transform of every individual row of the input
+        matrix; this flag enables you to transform multiple vectors simultaneously and can be used to
+        decrease the overhead (which is sometimes several times larger than the processing itself) to
+        perform 3D and higher-dimensional transformations and so forth.*/
+    DFT_ROWS           = 4,
+    /** performs a forward transformation of 1D or 2D real array; the result,
+        though being a complex array, has complex-conjugate symmetry (*CCS*, see the function
+        description below for details), and such an array can be packed into a real array of the same
+        size as input, which is the fastest option and which is what the function does by default;
+        however, you may wish to get a full complex array (for simpler spectrum analysis, and so on) -
+        pass the flag to enable the function to produce a full-size complex output array. */
+    DFT_COMPLEX_OUTPUT = 16,
+    /** performs an inverse transformation of a 1D or 2D complex array; the
+        result is normally a complex array of the same size, however, if the input array has
+        conjugate-complex symmetry (for example, it is a result of forward transformation with
+        DFT_COMPLEX_OUTPUT flag), the output is a real array; while the function itself does not
+        check whether the input is symmetrical or not, you can pass the flag and then the function
+        will assume the symmetry and produce the real output array (note that when the input is packed
+        into a real array and inverse transformation is executed, the function treats the input as a
+        packed complex-conjugate symmetrical array, and the output will also be a real array). */
+    DFT_REAL_OUTPUT    = 32,
+    /** specifies that input is complex input. If this flag is set, the input must have 2 channels.
+        On the other hand, for backwards compatibility reason, if input has 2 channels, input is
+        already considered complex. */
+    DFT_COMPLEX_INPUT  = 64,
+    /** performs an inverse 1D or 2D transform instead of the default forward transform. */
+    DCT_INVERSE        = DFT_INVERSE,
+    /** performs a forward or inverse transform of every individual row of the input
+        matrix. This flag enables you to transform multiple vectors simultaneously and can be used to
+        decrease the overhead (which is sometimes several times larger than the processing itself) to
+        perform 3D and higher-dimensional transforms and so forth.*/
+    DCT_ROWS           = DFT_ROWS
+};
+
+//! Various border types, image boundaries are denoted with `|`
+//! @see borderInterpolate, copyMakeBorder
+enum BorderTypes {
+    BORDER_CONSTANT    = 0, //!< `iiiiii|abcdefgh|iiiiiii`  with some specified `i`
+    BORDER_REPLICATE   = 1, //!< `aaaaaa|abcdefgh|hhhhhhh`
+    BORDER_REFLECT     = 2, //!< `fedcba|abcdefgh|hgfedcb`
+    BORDER_WRAP        = 3, //!< `cdefgh|abcdefgh|abcdefg`
+    BORDER_REFLECT_101 = 4, //!< `gfedcb|abcdefgh|gfedcba`
+    BORDER_TRANSPARENT = 5, //!< `uvwxyz|abcdefgh|ijklmno`
+
+    BORDER_REFLECT101  = BORDER_REFLECT_101, //!< same as BORDER_REFLECT_101
+    BORDER_DEFAULT     = BORDER_REFLECT_101, //!< same as BORDER_REFLECT_101
+    BORDER_ISOLATED    = 16 //!< do not look outside of ROI
+};
+
+//! @} core_array
+
+//! @addtogroup core_utils
+//! @{
+
+/*! @brief Signals an error and raises the exception.
+
+By default the function prints information about the error to stderr,
+then it either stops if setBreakOnError() had been called before or raises the exception.
+It is possible to alternate error processing by using redirectError().
+@param _code - error code (Error::Code)
+@param _err - error description
+@param _func - function name. Available only when the compiler supports getting it
+@param _file - source file name where the error has occurred
+@param _line - line number in the source file where the error has occurred
+@see CV_Error, CV_Error_, CV_Assert, CV_DbgAssert
+ */
+CV_EXPORTS void error(int _code, const String& _err, const char* _func, const char* _file, int _line);
+
+#ifdef __GNUC__
+# if defined __clang__ || defined __APPLE__
+#   pragma GCC diagnostic push
+#   pragma GCC diagnostic ignored "-Winvalid-noreturn"
+# endif
+#endif
+
+/** same as cv::error, but does not return */
+CV_INLINE CV_NORETURN void errorNoReturn(int _code, const String& _err, const char* _func, const char* _file, int _line)
+{
+    error(_code, _err, _func, _file, _line);
+#ifdef __GNUC__
+# if !defined __clang__ && !defined __APPLE__
+    // this suppresses this warning: "noreturn" function does return [enabled by default]
+    __builtin_trap();
+    // or use infinite loop: for (;;) {}
+# endif
+#endif
+}
+#ifdef __GNUC__
+# if defined __clang__ || defined __APPLE__
+#   pragma GCC diagnostic pop
+# endif
+#endif
+
+#ifdef CV_STATIC_ANALYSIS
+
+// In practice, some macro are not processed correctly (noreturn is not detected).
+// We need to use simplified definition for them.
+#define CV_Error(code, msg) do { (void)(code); (void)(msg); abort(); } while (0)
+#define CV_Error_(code, args) do { (void)(code); (void)(cv::format args); abort(); } while (0)
+#define CV_Assert( expr ) do { if (!(expr)) abort(); } while (0)
+#define CV_ErrorNoReturn CV_Error
+#define CV_ErrorNoReturn_ CV_Error_
+
+#else // CV_STATIC_ANALYSIS
+
+/** @brief Call the error handler.
+
+Currently, the error handler prints the error code and the error message to the standard
+error stream `stderr`. In the Debug configuration, it then provokes memory access violation, so that
+the execution stack and all the parameters can be analyzed by the debugger. In the Release
+configuration, the exception is thrown.
+
+@param code one of Error::Code
+@param msg error message
+*/
+#define CV_Error( code, msg ) cv::error( code, msg, CV_Func, __FILE__, __LINE__ )
+
+/**  @brief Call the error handler.
+
+This macro can be used to construct an error message on-fly to include some dynamic information,
+for example:
+@code
+    // note the extra parentheses around the formatted text message
+    CV_Error_(Error::StsOutOfRange,
+    ("the value at (%d, %d)=%g is out of range", badPt.x, badPt.y, badValue));
+@endcode
+@param code one of Error::Code
+@param args printf-like formatted error message in parentheses
+*/
+#define CV_Error_( code, args ) cv::error( code, cv::format args, CV_Func, __FILE__, __LINE__ )
+
+/** @brief Checks a condition at runtime and throws exception if it fails
+
+The macros CV_Assert (and CV_DbgAssert(expr)) evaluate the specified expression. If it is 0, the macros
+raise an error (see cv::error). The macro CV_Assert checks the condition in both Debug and Release
+configurations while CV_DbgAssert is only retained in the Debug configuration.
+*/
+#define CV_Assert( expr ) do { if(!!(expr)) ; else cv::error( cv::Error::StsAssert, #expr, CV_Func, __FILE__, __LINE__ ); } while(0)
+
+//! @cond IGNORED
+#define CV__ErrorNoReturn( code, msg ) cv::errorNoReturn( code, msg, CV_Func, __FILE__, __LINE__ )
+#define CV__ErrorNoReturn_( code, args ) cv::errorNoReturn( code, cv::format args, CV_Func, __FILE__, __LINE__ )
+#ifdef __OPENCV_BUILD
+#undef CV_Error
+#define CV_Error CV__ErrorNoReturn
+#undef CV_Error_
+#define CV_Error_ CV__ErrorNoReturn_
+#undef CV_Assert
+#define CV_Assert( expr ) do { if(!!(expr)) ; else cv::errorNoReturn( cv::Error::StsAssert, #expr, CV_Func, __FILE__, __LINE__ ); } while(0)
+#else
+// backward compatibility
+#define CV_ErrorNoReturn CV__ErrorNoReturn
+#define CV_ErrorNoReturn_ CV__ErrorNoReturn_
+#endif
+//! @endcond
+
+#endif // CV_STATIC_ANALYSIS
+
+//! @cond IGNORED
+
+#if defined OPENCV_FORCE_MULTIARG_ASSERT_CHECK && defined CV_STATIC_ANALYSIS
+#warning "OPENCV_FORCE_MULTIARG_ASSERT_CHECK can't be used with CV_STATIC_ANALYSIS"
+#undef OPENCV_FORCE_MULTIARG_ASSERT_CHECK
+#endif
+
+#ifdef OPENCV_FORCE_MULTIARG_ASSERT_CHECK
+#define CV_Assert_1( expr ) do { if(!!(expr)) ; else cv::error( cv::Error::StsAssert, #expr, CV_Func, __FILE__, __LINE__ ); } while(0)
+#else
+#define CV_Assert_1 CV_Assert
+#endif
+#define CV_Assert_2( expr1, expr2 ) CV_Assert_1(expr1); CV_Assert_1(expr2)
+#define CV_Assert_3( expr1, expr2, expr3 ) CV_Assert_2(expr1, expr2); CV_Assert_1(expr3)
+#define CV_Assert_4( expr1, expr2, expr3, expr4 ) CV_Assert_3(expr1, expr2, expr3); CV_Assert_1(expr4)
+#define CV_Assert_5( expr1, expr2, expr3, expr4, expr5 ) CV_Assert_4(expr1, expr2, expr3, expr4); CV_Assert_1(expr5)
+#define CV_Assert_6( expr1, expr2, expr3, expr4, expr5, expr6 ) CV_Assert_5(expr1, expr2, expr3, expr4, expr5); CV_Assert_1(expr6)
+#define CV_Assert_7( expr1, expr2, expr3, expr4, expr5, expr6, expr7 ) CV_Assert_6(expr1, expr2, expr3, expr4, expr5, expr6 ); CV_Assert_1(expr7)
+#define CV_Assert_8( expr1, expr2, expr3, expr4, expr5, expr6, expr7, expr8 ) CV_Assert_7(expr1, expr2, expr3, expr4, expr5, expr6, expr7 ); CV_Assert_1(expr8)
+#define CV_Assert_9( expr1, expr2, expr3, expr4, expr5, expr6, expr7, expr8, expr9 ) CV_Assert_8(expr1, expr2, expr3, expr4, expr5, expr6, expr7, expr8 ); CV_Assert_1(expr9)
+#define CV_Assert_10( expr1, expr2, expr3, expr4, expr5, expr6, expr7, expr8, expr9, expr10 ) CV_Assert_9(expr1, expr2, expr3, expr4, expr5, expr6, expr7, expr8, expr9 ); CV_Assert_1(expr10)
+
+#define CV_Assert_N(...) do { __CV_CAT(CV_Assert_, __CV_VA_NUM_ARGS(__VA_ARGS__)) (__VA_ARGS__); } while(0)
+
+#ifdef OPENCV_FORCE_MULTIARG_ASSERT_CHECK
+#undef CV_Assert
+#define CV_Assert CV_Assert_N
+#endif
+//! @endcond
+
+#if defined _DEBUG || defined CV_STATIC_ANALYSIS
+#  define CV_DbgAssert(expr) CV_Assert(expr)
+#else
+/** replaced with CV_Assert(expr) in Debug configuration */
+#  define CV_DbgAssert(expr)
+#endif
+
+/*
+ * Hamming distance functor - counts the bit differences between two strings - useful for the Brief descriptor
+ * bit count of A exclusive XOR'ed with B
+ */
+struct CV_EXPORTS Hamming
+{
+    enum { normType = NORM_HAMMING };
+    typedef unsigned char ValueType;
+    typedef int ResultType;
+
+    /** this will count the bits in a ^ b
+     */
+    ResultType operator()( const unsigned char* a, const unsigned char* b, int size ) const;
+};
+
+typedef Hamming HammingLUT;
+
+/////////////////////////////////// inline norms ////////////////////////////////////
+
+template<typename _Tp> inline _Tp cv_abs(_Tp x) { return std::abs(x); }
+inline int cv_abs(uchar x) { return x; }
+inline int cv_abs(schar x) { return std::abs(x); }
+inline int cv_abs(ushort x) { return x; }
+inline int cv_abs(short x) { return std::abs(x); }
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL2Sqr(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    int i=0;
+#if CV_ENABLE_UNROLLED
+    for( ; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = a[i], v1 = a[i+1], v2 = a[i+2], v3 = a[i+3];
+        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = a[i];
+        s += v*v;
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL1(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    int i = 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        s += (_AccTp)cv_abs(a[i]) + (_AccTp)cv_abs(a[i+1]) +
+            (_AccTp)cv_abs(a[i+2]) + (_AccTp)cv_abs(a[i+3]);
+    }
+#endif
+    for( ; i < n; i++ )
+        s += cv_abs(a[i]);
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normInf(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    for( int i = 0; i < n; i++ )
+        s = std::max(s, (_AccTp)cv_abs(a[i]));
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL2Sqr(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    int i= 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
+        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = _AccTp(a[i] - b[i]);
+        s += v*v;
+    }
+    return s;
+}
+
+static inline float normL2Sqr(const float* a, const float* b, int n)
+{
+    float s = 0.f;
+    for( int i = 0; i < n; i++ )
+    {
+        float v = a[i] - b[i];
+        s += v*v;
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL1(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    int i= 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
+        s += std::abs(v0) + std::abs(v1) + std::abs(v2) + std::abs(v3);
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = _AccTp(a[i] - b[i]);
+        s += std::abs(v);
+    }
+    return s;
+}
+
+inline float normL1(const float* a, const float* b, int n)
+{
+    float s = 0.f;
+    for( int i = 0; i < n; i++ )
+    {
+        s += std::abs(a[i] - b[i]);
+    }
+    return s;
+}
+
+inline int normL1(const uchar* a, const uchar* b, int n)
+{
+    int s = 0;
+    for( int i = 0; i < n; i++ )
+    {
+        s += std::abs(a[i] - b[i]);
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normInf(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    for( int i = 0; i < n; i++ )
+    {
+        _AccTp v0 = a[i] - b[i];
+        s = std::max(s, std::abs(v0));
+    }
+    return s;
+}
+
+/** @brief Computes the cube root of an argument.
+
+ The function cubeRoot computes \f$\sqrt[3]{\texttt{val}}\f$. Negative arguments are handled correctly.
+ NaN and Inf are not handled. The accuracy approaches the maximum possible accuracy for
+ single-precision data.
+ @param val A function argument.
+ */
+CV_EXPORTS_W float cubeRoot(float val);
+
+/** @overload
+
+cubeRoot with argument of `double` type calls `std::cbrt(double)` (C++11) or falls back on `pow()` for C++98 compilation mode.
+*/
+static inline
+double cubeRoot(double val)
+{
+#ifdef CV_CXX11
+    return std::cbrt(val);
+#else
+    double v = pow(abs(val), 1/3.);  // pow doesn't support negative inputs with fractional exponents
+    return val >= 0 ? v : -v;
+#endif
+}
+
+/** @brief Calculates the angle of a 2D vector in degrees.
+
+ The function fastAtan2 calculates the full-range angle of an input 2D vector. The angle is measured
+ in degrees and varies from 0 to 360 degrees. The accuracy is about 0.3 degrees.
+ @param x x-coordinate of the vector.
+ @param y y-coordinate of the vector.
+ */
+CV_EXPORTS_W float fastAtan2(float y, float x);
+
+/** proxy for hal::LU */
+CV_EXPORTS int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+/** proxy for hal::LU */
+CV_EXPORTS int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+/** proxy for hal::Cholesky */
+CV_EXPORTS bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+/** proxy for hal::Cholesky */
+CV_EXPORTS bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+
+////////////////// forward declarations for important OpenCV types //////////////////
+
+//! @cond IGNORED
+
+template<typename _Tp, int cn> class Vec;
+template<typename _Tp, int m, int n> class Matx;
+
+template<typename _Tp> class Complex;
+template<typename _Tp> class Point_;
+template<typename _Tp> class Point3_;
+template<typename _Tp> class Size_;
+template<typename _Tp> class Rect_;
+template<typename _Tp> class Scalar_;
+
+class CV_EXPORTS RotatedRect;
+class CV_EXPORTS Range;
+class CV_EXPORTS TermCriteria;
+class CV_EXPORTS KeyPoint;
+class CV_EXPORTS DMatch;
+class CV_EXPORTS RNG;
+
+class CV_EXPORTS Mat;
+class CV_EXPORTS MatExpr;
+
+class CV_EXPORTS UMat;
+
+class CV_EXPORTS SparseMat;
+typedef Mat MatND;
+
+template<typename _Tp> class Mat_;
+template<typename _Tp> class SparseMat_;
+
+class CV_EXPORTS MatConstIterator;
+class CV_EXPORTS SparseMatIterator;
+class CV_EXPORTS SparseMatConstIterator;
+template<typename _Tp> class MatIterator_;
+template<typename _Tp> class MatConstIterator_;
+template<typename _Tp> class SparseMatIterator_;
+template<typename _Tp> class SparseMatConstIterator_;
+
+namespace ogl
+{
+    class CV_EXPORTS Buffer;
+    class CV_EXPORTS Texture2D;
+    class CV_EXPORTS Arrays;
+}
+
+namespace cuda
+{
+    class CV_EXPORTS GpuMat;
+    class CV_EXPORTS HostMem;
+    class CV_EXPORTS Stream;
+    class CV_EXPORTS Event;
+}
+
+namespace cudev
+{
+    template <typename _Tp> class GpuMat_;
+}
+
+namespace ipp
+{
+#if OPENCV_ABI_COMPATIBILITY > 300
+CV_EXPORTS   unsigned long long getIppFeatures();
+#else
+CV_EXPORTS   int getIppFeatures();
+#endif
+CV_EXPORTS   void setIppStatus(int status, const char * const funcname = NULL, const char * const filename = NULL,
+                             int line = 0);
+CV_EXPORTS   int getIppStatus();
+CV_EXPORTS   String getIppErrorLocation();
+CV_EXPORTS_W bool   useIPP();
+CV_EXPORTS_W void   setUseIPP(bool flag);
+CV_EXPORTS_W String getIppVersion();
+
+// IPP Not-Exact mode. This function may force use of IPP then both IPP and OpenCV provide proper results
+// but have internal accuracy differences which have too much direct or indirect impact on accuracy tests.
+CV_EXPORTS_W bool useIPP_NotExact();
+CV_EXPORTS_W void setUseIPP_NotExact(bool flag);
+#if OPENCV_ABI_COMPATIBILITY < 400
+CV_EXPORTS_W bool useIPP_NE();
+CV_EXPORTS_W void setUseIPP_NE(bool flag);
+#endif
+
+} // ipp
+
+//! @endcond
+
+//! @} core_utils
+
+
+
+
+} // cv
+
+#include "opencv2/core/neon_utils.hpp"
+#include "opencv2/core/vsx_utils.hpp"
+#include "opencv2/core/check.hpp"
+
+#endif //OPENCV_CORE_BASE_HPP

3rdPart/include/opencv2/core/bindings_utils.hpp

@@ -0,0 +1,170 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_BINDINGS_UTILS_HPP
+#define OPENCV_CORE_BINDINGS_UTILS_HPP
+
+#include <opencv2/core/async.hpp>
+#include <opencv2/core/detail/async_promise.hpp>
+#include <opencv2/core/utils/logger.hpp>
+
+#include <stdexcept>
+
+namespace cv { namespace utils {
+//! @addtogroup core_utils
+//! @{
+
+CV_EXPORTS_W String dumpInputArray(InputArray argument);
+
+CV_EXPORTS_W String dumpInputArrayOfArrays(InputArrayOfArrays argument);
+
+CV_EXPORTS_W String dumpInputOutputArray(InputOutputArray argument);
+
+CV_EXPORTS_W String dumpInputOutputArrayOfArrays(InputOutputArrayOfArrays argument);
+
+CV_WRAP static inline
+String dumpBool(bool argument)
+{
+    return (argument) ? String("Bool: True") : String("Bool: False");
+}
+
+CV_WRAP static inline
+String dumpInt(int argument)
+{
+    return cv::format("Int: %d", argument);
+}
+
+CV_WRAP static inline
+String dumpSizeT(size_t argument)
+{
+    std::ostringstream oss("size_t: ", std::ios::ate);
+    oss << argument;
+    return oss.str();
+}
+
+CV_WRAP static inline
+String dumpFloat(float argument)
+{
+    return cv::format("Float: %.2f", argument);
+}
+
+CV_WRAP static inline
+String dumpDouble(double argument)
+{
+    return cv::format("Double: %.2f", argument);
+}
+
+CV_WRAP static inline
+String dumpCString(const char* argument)
+{
+    return cv::format("String: %s", argument);
+}
+
+CV_WRAP static inline
+String dumpString(const String& argument)
+{
+    return cv::format("String: %s", argument.c_str());
+}
+
+CV_WRAP static inline
+String testOverloadResolution(int value, const Point& point = Point(42, 24))
+{
+    return format("overload (int=%d, point=(x=%d, y=%d))", value, point.x,
+                  point.y);
+}
+
+CV_WRAP static inline
+String testOverloadResolution(const Rect& rect)
+{
+    return format("overload (rect=(x=%d, y=%d, w=%d, h=%d))", rect.x, rect.y,
+                  rect.width, rect.height);
+}
+
+CV_WRAP static inline
+String dumpRect(const Rect& argument)
+{
+    return format("rect: (x=%d, y=%d, w=%d, h=%d)", argument.x, argument.y,
+                  argument.width, argument.height);
+}
+
+CV_WRAP static inline
+String dumpTermCriteria(const TermCriteria& argument)
+{
+    return format("term_criteria: (type=%d, max_count=%d, epsilon=%lf",
+                  argument.type, argument.maxCount, argument.epsilon);
+}
+
+CV_WRAP static inline
+String dumpRotatedRect(const RotatedRect& argument)
+{
+    return format("rotated_rect: (c_x=%f, c_y=%f, w=%f, h=%f, a=%f)",
+                  argument.center.x, argument.center.y, argument.size.width,
+                  argument.size.height, argument.angle);
+}
+
+CV_WRAP static inline
+String dumpRange(const Range& argument)
+{
+    if (argument == Range::all())
+    {
+        return "range: all";
+    }
+    else
+    {
+        return format("range: (s=%d, e=%d)", argument.start, argument.end);
+    }
+}
+
+CV_WRAP static inline
+void testRaiseGeneralException()
+{
+    throw std::runtime_error("exception text");
+}
+
+CV_WRAP static inline
+AsyncArray testAsyncArray(InputArray argument)
+{
+    AsyncPromise p;
+    p.setValue(argument);
+    return p.getArrayResult();
+}
+
+CV_WRAP static inline
+AsyncArray testAsyncException()
+{
+    AsyncPromise p;
+    try
+    {
+        CV_Error(Error::StsOk, "Test: Generated async error");
+    }
+    catch (const cv::Exception& e)
+    {
+        p.setException(e);
+    }
+    return p.getArrayResult();
+}
+
+//! @}  // core_utils
+}  // namespace cv::utils
+
+//! @cond IGNORED
+
+CV_WRAP static inline
+int setLogLevel(int level)
+{
+    // NB: Binding generators doesn't work with enums properly yet, so we define separate overload here
+    return cv::utils::logging::setLogLevel((cv::utils::logging::LogLevel)level);
+}
+
+CV_WRAP static inline
+int getLogLevel()
+{
+    return cv::utils::logging::getLogLevel();
+}
+
+//! @endcond IGNORED
+
+} // namespaces cv /  utils
+
+#endif // OPENCV_CORE_BINDINGS_UTILS_HPP

3rdPart/include/opencv2/core/bufferpool.hpp

@@ -0,0 +1,40 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+//
+// Copyright (C) 2014, Advanced Micro Devices, Inc., all rights reserved.
+
+#ifndef OPENCV_CORE_BUFFER_POOL_HPP
+#define OPENCV_CORE_BUFFER_POOL_HPP
+
+#ifdef _MSC_VER
+#pragma warning(push)
+#pragma warning(disable: 4265)
+#endif
+
+namespace cv
+{
+
+//! @addtogroup core
+//! @{
+
+class BufferPoolController
+{
+protected:
+    ~BufferPoolController() { }
+public:
+    virtual size_t getReservedSize() const = 0;
+    virtual size_t getMaxReservedSize() const = 0;
+    virtual void setMaxReservedSize(size_t size) = 0;
+    virtual void freeAllReservedBuffers() = 0;
+};
+
+//! @}
+
+}
+
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif
+
+#endif // OPENCV_CORE_BUFFER_POOL_HPP

3rdPart/include/opencv2/core/check.hpp

@@ -0,0 +1,160 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_CHECK_HPP
+#define OPENCV_CORE_CHECK_HPP
+
+#include <opencv2/core/base.hpp>
+
+namespace cv {
+
+/** Returns string of cv::Mat depth value: CV_8U -> "CV_8U" or "<invalid depth>" */
+CV_EXPORTS const char* depthToString(int depth);
+
+/** Returns string of cv::Mat depth value: CV_8UC3 -> "CV_8UC3" or "<invalid type>" */
+CV_EXPORTS const String typeToString(int type);
+
+
+//! @cond IGNORED
+namespace detail {
+
+/** Returns string of cv::Mat depth value: CV_8U -> "CV_8U" or NULL */
+CV_EXPORTS const char* depthToString_(int depth);
+
+/** Returns string of cv::Mat depth value: CV_8UC3 -> "CV_8UC3" or cv::String() */
+CV_EXPORTS const cv::String typeToString_(int type);
+
+enum TestOp {
+  TEST_CUSTOM = 0,
+  TEST_EQ = 1,
+  TEST_NE = 2,
+  TEST_LE = 3,
+  TEST_LT = 4,
+  TEST_GE = 5,
+  TEST_GT = 6,
+  CV__LAST_TEST_OP
+};
+
+struct CheckContext {
+    const char* func;
+    const char* file;
+    int line;
+    enum TestOp testOp;
+    const char* message;
+    const char* p1_str;
+    const char* p2_str;
+};
+
+#ifndef CV__CHECK_FILENAME
+# define CV__CHECK_FILENAME __FILE__
+#endif
+
+#ifndef CV__CHECK_FUNCTION
+# if defined _MSC_VER
+#   define CV__CHECK_FUNCTION __FUNCSIG__
+# elif defined __GNUC__
+#   define CV__CHECK_FUNCTION __PRETTY_FUNCTION__
+# else
+#   define CV__CHECK_FUNCTION "<unknown>"
+# endif
+#endif
+
+#define CV__CHECK_LOCATION_VARNAME(id) CVAUX_CONCAT(CVAUX_CONCAT(__cv_check_, id), __LINE__)
+#define CV__DEFINE_CHECK_CONTEXT(id, message, testOp, p1_str, p2_str) \
+    static const cv::detail::CheckContext CV__CHECK_LOCATION_VARNAME(id) = \
+            { CV__CHECK_FUNCTION, CV__CHECK_FILENAME, __LINE__, testOp, "" message, "" p1_str, "" p2_str }
+
+CV_EXPORTS void CV_NORETURN check_failed_auto(const int v1, const int v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const size_t v1, const size_t v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const float v1, const float v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const double v1, const double v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const Size_<int> v1, const Size_<int> v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatDepth(const int v1, const int v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatType(const int v1, const int v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatChannels(const int v1, const int v2, const CheckContext& ctx);
+
+CV_EXPORTS void CV_NORETURN check_failed_auto(const int v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const size_t v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const float v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const double v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const Size_<int> v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const std::string& v1, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatDepth(const int v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatType(const int v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatChannels(const int v, const CheckContext& ctx);
+
+
+#define CV__TEST_EQ(v1, v2) ((v1) == (v2))
+#define CV__TEST_NE(v1, v2) ((v1) != (v2))
+#define CV__TEST_LE(v1, v2) ((v1) <= (v2))
+#define CV__TEST_LT(v1, v2) ((v1) < (v2))
+#define CV__TEST_GE(v1, v2) ((v1) >= (v2))
+#define CV__TEST_GT(v1, v2) ((v1) > (v2))
+
+#define CV__CHECK(id, op, type, v1, v2, v1_str, v2_str, msg_str) do { \
+    if(CV__TEST_##op((v1), (v2))) ; else { \
+        CV__DEFINE_CHECK_CONTEXT(id, msg_str, cv::detail::TEST_ ## op, v1_str, v2_str); \
+        cv::detail::check_failed_ ## type((v1), (v2), CV__CHECK_LOCATION_VARNAME(id)); \
+    } \
+} while (0)
+
+#define CV__CHECK_CUSTOM_TEST(id, type, v, test_expr, v_str, test_expr_str, msg_str) do { \
+    if(!!(test_expr)) ; else { \
+        CV__DEFINE_CHECK_CONTEXT(id, msg_str, cv::detail::TEST_CUSTOM, v_str, test_expr_str); \
+        cv::detail::check_failed_ ## type((v), CV__CHECK_LOCATION_VARNAME(id)); \
+    } \
+} while (0)
+
+} // namespace
+//! @endcond
+
+
+/// Supported values of these types: int, float, double
+#define CV_CheckEQ(v1, v2, msg)  CV__CHECK(_, EQ, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckNE(v1, v2, msg)  CV__CHECK(_, NE, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckLE(v1, v2, msg)  CV__CHECK(_, LE, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckLT(v1, v2, msg)  CV__CHECK(_, LT, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckGE(v1, v2, msg)  CV__CHECK(_, GE, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckGT(v1, v2, msg)  CV__CHECK(_, GT, auto, v1, v2, #v1, #v2, msg)
+
+/// Check with additional "decoding" of type values in error message
+#define CV_CheckTypeEQ(t1, t2, msg)  CV__CHECK(_, EQ, MatType, t1, t2, #t1, #t2, msg)
+/// Check with additional "decoding" of depth values in error message
+#define CV_CheckDepthEQ(d1, d2, msg)  CV__CHECK(_, EQ, MatDepth, d1, d2, #d1, #d2, msg)
+
+#define CV_CheckChannelsEQ(c1, c2, msg)  CV__CHECK(_, EQ, MatChannels, c1, c2, #c1, #c2, msg)
+
+/// Example: type == CV_8UC1 || type == CV_8UC3
+#define CV_CheckType(t, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, MatType, t, (test_expr), #t, #test_expr, msg)
+
+/// Example: depth == CV_32F || depth == CV_64F
+#define CV_CheckDepth(t, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, MatDepth, t, (test_expr), #t, #test_expr, msg)
+
+/// Example: v == A || v == B
+#define CV_Check(v, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, auto, v, (test_expr), #v, #test_expr, msg)
+
+/// Some complex conditions: CV_Check(src2, src2.empty() || (src2.type() == src1.type() && src2.size() == src1.size()), "src2 should have same size/type as src1")
+// TODO define pretty-printers
+
+#ifndef NDEBUG
+#define CV_DbgCheck(v, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, auto, v, (test_expr), #v, #test_expr, msg)
+#define CV_DbgCheckEQ(v1, v2, msg)  CV__CHECK(_, EQ, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckNE(v1, v2, msg)  CV__CHECK(_, NE, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckLE(v1, v2, msg)  CV__CHECK(_, LE, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckLT(v1, v2, msg)  CV__CHECK(_, LT, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckGE(v1, v2, msg)  CV__CHECK(_, GE, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckGT(v1, v2, msg)  CV__CHECK(_, GT, auto, v1, v2, #v1, #v2, msg)
+#else
+#define CV_DbgCheck(v, test_expr, msg)  do { } while (0)
+#define CV_DbgCheckEQ(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckNE(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckLE(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckLT(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckGE(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckGT(v1, v2, msg)  do { } while (0)
+#endif
+
+} // namespace
+
+#endif // OPENCV_CORE_CHECK_HPP

3rdPart/include/opencv2/core/core.hpp

@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __OPENCV_BUILD
+#error this is a compatibility header which should not be used inside the OpenCV library
+#endif
+
+#include "opencv2/core.hpp"

3rdPart/include/opencv2/core/cuda.inl.hpp

@@ -0,0 +1,631 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CUDAINL_HPP
+#define OPENCV_CORE_CUDAINL_HPP
+
+#include "opencv2/core/cuda.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda {
+
+//===================================================================================
+// GpuMat
+//===================================================================================
+
+inline
+GpuMat::GpuMat(Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{}
+
+inline
+GpuMat::GpuMat(int rows_, int cols_, int type_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (rows_ > 0 && cols_ > 0)
+        create(rows_, cols_, type_);
+}
+
+inline
+GpuMat::GpuMat(Size size_, int type_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (size_.height > 0 && size_.width > 0)
+        create(size_.height, size_.width, type_);
+}
+
+inline
+GpuMat::GpuMat(int rows_, int cols_, int type_, Scalar s_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (rows_ > 0 && cols_ > 0)
+    {
+        create(rows_, cols_, type_);
+        setTo(s_);
+    }
+}
+
+inline
+GpuMat::GpuMat(Size size_, int type_, Scalar s_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (size_.height > 0 && size_.width > 0)
+    {
+        create(size_.height, size_.width, type_);
+        setTo(s_);
+    }
+}
+
+inline
+GpuMat::GpuMat(const GpuMat& m)
+    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend), allocator(m.allocator)
+{
+    if (refcount)
+        CV_XADD(refcount, 1);
+}
+
+inline
+GpuMat::GpuMat(InputArray arr, Allocator* allocator_) :
+    flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    upload(arr);
+}
+
+inline
+GpuMat::~GpuMat()
+{
+    release();
+}
+
+inline
+GpuMat& GpuMat::operator =(const GpuMat& m)
+{
+    if (this != &m)
+    {
+        GpuMat temp(m);
+        swap(temp);
+    }
+
+    return *this;
+}
+
+inline
+void GpuMat::create(Size size_, int type_)
+{
+    create(size_.height, size_.width, type_);
+}
+
+inline
+void GpuMat::swap(GpuMat& b)
+{
+    std::swap(flags, b.flags);
+    std::swap(rows, b.rows);
+    std::swap(cols, b.cols);
+    std::swap(step, b.step);
+    std::swap(data, b.data);
+    std::swap(datastart, b.datastart);
+    std::swap(dataend, b.dataend);
+    std::swap(refcount, b.refcount);
+    std::swap(allocator, b.allocator);
+}
+
+inline
+GpuMat GpuMat::clone() const
+{
+    GpuMat m;
+    copyTo(m);
+    return m;
+}
+
+inline
+void GpuMat::copyTo(OutputArray dst, InputArray mask) const
+{
+    copyTo(dst, mask, Stream::Null());
+}
+
+inline
+GpuMat& GpuMat::setTo(Scalar s)
+{
+    return setTo(s, Stream::Null());
+}
+
+inline
+GpuMat& GpuMat::setTo(Scalar s, InputArray mask)
+{
+    return setTo(s, mask, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype) const
+{
+    convertTo(dst, rtype, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype, double alpha, double beta) const
+{
+    convertTo(dst, rtype, alpha, beta, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype, double alpha, Stream& stream) const
+{
+    convertTo(dst, rtype, alpha, 0.0, stream);
+}
+
+inline
+void GpuMat::assignTo(GpuMat& m, int _type) const
+{
+    if (_type < 0)
+        m = *this;
+    else
+        convertTo(m, _type);
+}
+
+inline
+uchar* GpuMat::ptr(int y)
+{
+    CV_DbgAssert( (unsigned)y < (unsigned)rows );
+    return data + step * y;
+}
+
+inline
+const uchar* GpuMat::ptr(int y) const
+{
+    CV_DbgAssert( (unsigned)y < (unsigned)rows );
+    return data + step * y;
+}
+
+template<typename _Tp> inline
+_Tp* GpuMat::ptr(int y)
+{
+    return (_Tp*)ptr(y);
+}
+
+template<typename _Tp> inline
+const _Tp* GpuMat::ptr(int y) const
+{
+    return (const _Tp*)ptr(y);
+}
+
+template <class T> inline
+GpuMat::operator PtrStepSz<T>() const
+{
+    return PtrStepSz<T>(rows, cols, (T*)data, step);
+}
+
+template <class T> inline
+GpuMat::operator PtrStep<T>() const
+{
+    return PtrStep<T>((T*)data, step);
+}
+
+inline
+GpuMat GpuMat::row(int y) const
+{
+    return GpuMat(*this, Range(y, y+1), Range::all());
+}
+
+inline
+GpuMat GpuMat::col(int x) const
+{
+    return GpuMat(*this, Range::all(), Range(x, x+1));
+}
+
+inline
+GpuMat GpuMat::rowRange(int startrow, int endrow) const
+{
+    return GpuMat(*this, Range(startrow, endrow), Range::all());
+}
+
+inline
+GpuMat GpuMat::rowRange(Range r) const
+{
+    return GpuMat(*this, r, Range::all());
+}
+
+inline
+GpuMat GpuMat::colRange(int startcol, int endcol) const
+{
+    return GpuMat(*this, Range::all(), Range(startcol, endcol));
+}
+
+inline
+GpuMat GpuMat::colRange(Range r) const
+{
+    return GpuMat(*this, Range::all(), r);
+}
+
+inline
+GpuMat GpuMat::operator ()(Range rowRange_, Range colRange_) const
+{
+    return GpuMat(*this, rowRange_, colRange_);
+}
+
+inline
+GpuMat GpuMat::operator ()(Rect roi) const
+{
+    return GpuMat(*this, roi);
+}
+
+inline
+bool GpuMat::isContinuous() const
+{
+    return (flags & Mat::CONTINUOUS_FLAG) != 0;
+}
+
+inline
+size_t GpuMat::elemSize() const
+{
+    return CV_ELEM_SIZE(flags);
+}
+
+inline
+size_t GpuMat::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int GpuMat::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int GpuMat::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int GpuMat::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+size_t GpuMat::step1() const
+{
+    return step / elemSize1();
+}
+
+inline
+Size GpuMat::size() const
+{
+    return Size(cols, rows);
+}
+
+inline
+bool GpuMat::empty() const
+{
+    return data == 0;
+}
+
+static inline
+GpuMat createContinuous(int rows, int cols, int type)
+{
+    GpuMat m;
+    createContinuous(rows, cols, type, m);
+    return m;
+}
+
+static inline
+void createContinuous(Size size, int type, OutputArray arr)
+{
+    createContinuous(size.height, size.width, type, arr);
+}
+
+static inline
+GpuMat createContinuous(Size size, int type)
+{
+    GpuMat m;
+    createContinuous(size, type, m);
+    return m;
+}
+
+static inline
+void ensureSizeIsEnough(Size size, int type, OutputArray arr)
+{
+    ensureSizeIsEnough(size.height, size.width, type, arr);
+}
+
+static inline
+void swap(GpuMat& a, GpuMat& b)
+{
+    a.swap(b);
+}
+
+//===================================================================================
+// HostMem
+//===================================================================================
+
+inline
+HostMem::HostMem(AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+}
+
+inline
+HostMem::HostMem(const HostMem& m)
+    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend), alloc_type(m.alloc_type)
+{
+    if( refcount )
+        CV_XADD(refcount, 1);
+}
+
+inline
+HostMem::HostMem(int rows_, int cols_, int type_, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    if (rows_ > 0 && cols_ > 0)
+        create(rows_, cols_, type_);
+}
+
+inline
+HostMem::HostMem(Size size_, int type_, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    if (size_.height > 0 && size_.width > 0)
+        create(size_.height, size_.width, type_);
+}
+
+inline
+HostMem::HostMem(InputArray arr, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    arr.getMat().copyTo(*this);
+}
+
+inline
+HostMem::~HostMem()
+{
+    release();
+}
+
+inline
+HostMem& HostMem::operator =(const HostMem& m)
+{
+    if (this != &m)
+    {
+        HostMem temp(m);
+        swap(temp);
+    }
+
+    return *this;
+}
+
+inline
+void HostMem::swap(HostMem& b)
+{
+    std::swap(flags, b.flags);
+    std::swap(rows, b.rows);
+    std::swap(cols, b.cols);
+    std::swap(step, b.step);
+    std::swap(data, b.data);
+    std::swap(datastart, b.datastart);
+    std::swap(dataend, b.dataend);
+    std::swap(refcount, b.refcount);
+    std::swap(alloc_type, b.alloc_type);
+}
+
+inline
+HostMem HostMem::clone() const
+{
+    HostMem m(size(), type(), alloc_type);
+    createMatHeader().copyTo(m);
+    return m;
+}
+
+inline
+void HostMem::create(Size size_, int type_)
+{
+    create(size_.height, size_.width, type_);
+}
+
+inline
+Mat HostMem::createMatHeader() const
+{
+    return Mat(size(), type(), data, step);
+}
+
+inline
+bool HostMem::isContinuous() const
+{
+    return (flags & Mat::CONTINUOUS_FLAG) != 0;
+}
+
+inline
+size_t HostMem::elemSize() const
+{
+    return CV_ELEM_SIZE(flags);
+}
+
+inline
+size_t HostMem::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int HostMem::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int HostMem::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int HostMem::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+size_t HostMem::step1() const
+{
+    return step / elemSize1();
+}
+
+inline
+Size HostMem::size() const
+{
+    return Size(cols, rows);
+}
+
+inline
+bool HostMem::empty() const
+{
+    return data == 0;
+}
+
+static inline
+void swap(HostMem& a, HostMem& b)
+{
+    a.swap(b);
+}
+
+//===================================================================================
+// Stream
+//===================================================================================
+
+inline
+Stream::Stream(const Ptr<Impl>& impl)
+    : impl_(impl)
+{
+}
+
+//===================================================================================
+// Event
+//===================================================================================
+
+inline
+Event::Event(const Ptr<Impl>& impl)
+    : impl_(impl)
+{
+}
+
+//===================================================================================
+// Initialization & Info
+//===================================================================================
+
+inline
+bool TargetArchs::has(int major, int minor)
+{
+    return hasPtx(major, minor) || hasBin(major, minor);
+}
+
+inline
+bool TargetArchs::hasEqualOrGreater(int major, int minor)
+{
+    return hasEqualOrGreaterPtx(major, minor) || hasEqualOrGreaterBin(major, minor);
+}
+
+inline
+DeviceInfo::DeviceInfo()
+{
+    device_id_ = getDevice();
+}
+
+inline
+DeviceInfo::DeviceInfo(int device_id)
+{
+    CV_Assert( device_id >= 0 && device_id < getCudaEnabledDeviceCount() );
+    device_id_ = device_id;
+}
+
+inline
+int DeviceInfo::deviceID() const
+{
+    return device_id_;
+}
+
+inline
+size_t DeviceInfo::freeMemory() const
+{
+    size_t _totalMemory = 0, _freeMemory = 0;
+    queryMemory(_totalMemory, _freeMemory);
+    return _freeMemory;
+}
+
+inline
+size_t DeviceInfo::totalMemory() const
+{
+    size_t _totalMemory = 0, _freeMemory = 0;
+    queryMemory(_totalMemory, _freeMemory);
+    return _totalMemory;
+}
+
+inline
+bool DeviceInfo::supports(FeatureSet feature_set) const
+{
+    int version = majorVersion() * 10 + minorVersion();
+    return version >= feature_set;
+}
+
+
+}} // namespace cv { namespace cuda {
+
+//===================================================================================
+// Mat
+//===================================================================================
+
+namespace cv {
+
+inline
+Mat::Mat(const cuda::GpuMat& m)
+    : flags(0), dims(0), rows(0), cols(0), data(0), datastart(0), dataend(0), datalimit(0), allocator(0), u(0), size(&rows)
+{
+    m.download(*this);
+}
+
+}
+
+//! @endcond
+
+#endif // OPENCV_CORE_CUDAINL_HPP

3rdPart/include/opencv2/core/cuda/block.hpp

@@ -0,0 +1,211 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DEVICE_BLOCK_HPP
+#define OPENCV_CUDA_DEVICE_BLOCK_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct Block
+    {
+        static __device__ __forceinline__ unsigned int id()
+        {
+            return blockIdx.x;
+        }
+
+        static __device__ __forceinline__ unsigned int stride()
+        {
+            return blockDim.x * blockDim.y * blockDim.z;
+        }
+
+        static __device__ __forceinline__ void sync()
+        {
+            __syncthreads();
+        }
+
+        static __device__ __forceinline__ int flattenedThreadId()
+        {
+            return threadIdx.z * blockDim.x * blockDim.y + threadIdx.y * blockDim.x + threadIdx.x;
+        }
+
+        template<typename It, typename T>
+        static __device__ __forceinline__ void fill(It beg, It end, const T& value)
+        {
+            int STRIDE = stride();
+            It t = beg + flattenedThreadId();
+
+            for(; t < end; t += STRIDE)
+                *t = value;
+        }
+
+        template<typename OutIt, typename T>
+        static __device__ __forceinline__ void yota(OutIt beg, OutIt end, T value)
+        {
+            int STRIDE = stride();
+            int tid = flattenedThreadId();
+            value += tid;
+
+            for(OutIt t = beg + tid; t < end; t += STRIDE, value += STRIDE)
+                *t = value;
+        }
+
+        template<typename InIt, typename OutIt>
+        static __device__ __forceinline__ void copy(InIt beg, InIt end, OutIt out)
+        {
+            int STRIDE = stride();
+            InIt  t = beg + flattenedThreadId();
+            OutIt o = out + (t - beg);
+
+            for(; t < end; t += STRIDE, o += STRIDE)
+                *o = *t;
+        }
+
+        template<typename InIt, typename OutIt, class UnOp>
+        static __device__ __forceinline__ void transform(InIt beg, InIt end, OutIt out, UnOp op)
+        {
+            int STRIDE = stride();
+            InIt  t = beg + flattenedThreadId();
+            OutIt o = out + (t - beg);
+
+            for(; t < end; t += STRIDE, o += STRIDE)
+                *o = op(*t);
+        }
+
+        template<typename InIt1, typename InIt2, typename OutIt, class BinOp>
+        static __device__ __forceinline__ void transform(InIt1 beg1, InIt1 end1, InIt2 beg2, OutIt out, BinOp op)
+        {
+            int STRIDE = stride();
+            InIt1 t1 = beg1 + flattenedThreadId();
+            InIt2 t2 = beg2 + flattenedThreadId();
+            OutIt o  = out + (t1 - beg1);
+
+            for(; t1 < end1; t1 += STRIDE, t2 += STRIDE, o += STRIDE)
+                *o = op(*t1, *t2);
+        }
+
+        template<int CTA_SIZE, typename T, class BinOp>
+        static __device__ __forceinline__ void reduce(volatile T* buffer, BinOp op)
+        {
+            int tid = flattenedThreadId();
+            T val =  buffer[tid];
+
+            if (CTA_SIZE >= 1024) { if (tid < 512) buffer[tid] = val = op(val, buffer[tid + 512]); __syncthreads(); }
+            if (CTA_SIZE >=  512) { if (tid < 256) buffer[tid] = val = op(val, buffer[tid + 256]); __syncthreads(); }
+            if (CTA_SIZE >=  256) { if (tid < 128) buffer[tid] = val = op(val, buffer[tid + 128]); __syncthreads(); }
+            if (CTA_SIZE >=  128) { if (tid <  64) buffer[tid] = val = op(val, buffer[tid +  64]); __syncthreads(); }
+
+            if (tid < 32)
+            {
+                if (CTA_SIZE >=   64) { buffer[tid] = val = op(val, buffer[tid +  32]); }
+                if (CTA_SIZE >=   32) { buffer[tid] = val = op(val, buffer[tid +  16]); }
+                if (CTA_SIZE >=   16) { buffer[tid] = val = op(val, buffer[tid +   8]); }
+                if (CTA_SIZE >=    8) { buffer[tid] = val = op(val, buffer[tid +   4]); }
+                if (CTA_SIZE >=    4) { buffer[tid] = val = op(val, buffer[tid +   2]); }
+                if (CTA_SIZE >=    2) { buffer[tid] = val = op(val, buffer[tid +   1]); }
+            }
+        }
+
+        template<int CTA_SIZE, typename T, class BinOp>
+        static __device__ __forceinline__ T reduce(volatile T* buffer, T init, BinOp op)
+        {
+            int tid = flattenedThreadId();
+            T val =  buffer[tid] = init;
+            __syncthreads();
+
+            if (CTA_SIZE >= 1024) { if (tid < 512) buffer[tid] = val = op(val, buffer[tid + 512]); __syncthreads(); }
+            if (CTA_SIZE >=  512) { if (tid < 256) buffer[tid] = val = op(val, buffer[tid + 256]); __syncthreads(); }
+            if (CTA_SIZE >=  256) { if (tid < 128) buffer[tid] = val = op(val, buffer[tid + 128]); __syncthreads(); }
+            if (CTA_SIZE >=  128) { if (tid <  64) buffer[tid] = val = op(val, buffer[tid +  64]); __syncthreads(); }
+
+            if (tid < 32)
+            {
+                if (CTA_SIZE >=   64) { buffer[tid] = val = op(val, buffer[tid +  32]); }
+                if (CTA_SIZE >=   32) { buffer[tid] = val = op(val, buffer[tid +  16]); }
+                if (CTA_SIZE >=   16) { buffer[tid] = val = op(val, buffer[tid +   8]); }
+                if (CTA_SIZE >=    8) { buffer[tid] = val = op(val, buffer[tid +   4]); }
+                if (CTA_SIZE >=    4) { buffer[tid] = val = op(val, buffer[tid +   2]); }
+                if (CTA_SIZE >=    2) { buffer[tid] = val = op(val, buffer[tid +   1]); }
+            }
+            __syncthreads();
+            return buffer[0];
+        }
+
+        template <typename T, class BinOp>
+        static __device__ __forceinline__ void reduce_n(T* data, unsigned int n, BinOp op)
+        {
+            int ftid = flattenedThreadId();
+            int sft = stride();
+
+            if (sft < n)
+            {
+                for (unsigned int i = sft + ftid; i < n; i += sft)
+                    data[ftid] = op(data[ftid], data[i]);
+
+                __syncthreads();
+
+                n = sft;
+            }
+
+            while (n > 1)
+            {
+                unsigned int half = n/2;
+
+                if (ftid < half)
+                    data[ftid] = op(data[ftid], data[n - ftid - 1]);
+
+                __syncthreads();
+
+                n = n - half;
+            }
+        }
+    };
+}}}
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_DEVICE_BLOCK_HPP */

3rdPart/include/opencv2/core/cuda/border_interpolate.hpp

@@ -0,0 +1,722 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_BORDER_INTERPOLATE_HPP
+#define OPENCV_CUDA_BORDER_INTERPOLATE_HPP
+
+#include "saturate_cast.hpp"
+#include "vec_traits.hpp"
+#include "vec_math.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    //////////////////////////////////////////////////////////////
+    // BrdConstant
+
+    template <typename D> struct BrdRowConstant
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowConstant(int width_, const D& val_ = VecTraits<D>::all(0)) : width(width_), val(val_) {}
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return x >= 0 ? saturate_cast<D>(data[x]) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return x < width ? saturate_cast<D>(data[x]) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return (x >= 0 && x < width) ? saturate_cast<D>(data[x]) : val;
+        }
+
+        int width;
+        D val;
+    };
+
+    template <typename D> struct BrdColConstant
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColConstant(int height_, const D& val_ = VecTraits<D>::all(0)) : height(height_), val(val_) {}
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return y >= 0 ? saturate_cast<D>(*(const T*)((const char*)data + y * step)) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return y < height ? saturate_cast<D>(*(const T*)((const char*)data + y * step)) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return (y >= 0 && y < height) ? saturate_cast<D>(*(const T*)((const char*)data + y * step)) : val;
+        }
+
+        int height;
+        D val;
+    };
+
+    template <typename D> struct BrdConstant
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdConstant(int height_, int width_, const D& val_ = VecTraits<D>::all(0)) : height(height_), width(width_), val(val_)
+        {
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(((const T*)((const uchar*)data + y * step))[x]) : val;
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(src(y, x)) : val;
+        }
+
+        int height;
+        int width;
+        D val;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdReplicate
+
+    template <typename D> struct BrdRowReplicate
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowReplicate(int width) : last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowReplicate(int width, U) : last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::max(x, 0);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::min(x, last_col);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int last_col;
+    };
+
+    template <typename D> struct BrdColReplicate
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColReplicate(int height) : last_row(height - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColReplicate(int height, U) : last_row(height - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::max(y, 0);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::min(y, last_row);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const T*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const T*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const T*)((const char*)data + idx_row(y) * step));
+        }
+
+        int last_row;
+    };
+
+    template <typename D> struct BrdReplicate
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdReplicate(int height, int width) : last_row(height - 1), last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdReplicate(int height, int width, U) : last_row(height - 1), last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::max(y, 0);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::min(y, last_row);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::max(x, 0);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::min(x, last_col);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int last_row;
+        int last_col;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdReflect101
+
+    template <typename D> struct BrdRowReflect101
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowReflect101(int width) : last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowReflect101(int width, U) : last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::abs(x) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::abs(last_col - ::abs(last_col - x)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int last_col;
+    };
+
+    template <typename D> struct BrdColReflect101
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColReflect101(int height) : last_row(height - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColReflect101(int height, U) : last_row(height - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::abs(y) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::abs(last_row - ::abs(last_row - y)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
+        }
+
+        int last_row;
+    };
+
+    template <typename D> struct BrdReflect101
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdReflect101(int height, int width) : last_row(height - 1), last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdReflect101(int height, int width, U) : last_row(height - 1), last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::abs(y) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::abs(last_row - ::abs(last_row - y)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::abs(x) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::abs(last_col - ::abs(last_col - x)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int last_row;
+        int last_col;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdReflect
+
+    template <typename D> struct BrdRowReflect
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowReflect(int width) : last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowReflect(int width, U) : last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (::abs(x) - (x < 0)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::abs(last_col - ::abs(last_col - x) + (x > last_col)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_high(::abs(x) - (x < 0));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int last_col;
+    };
+
+    template <typename D> struct BrdColReflect
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColReflect(int height) : last_row(height - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColReflect(int height, U) : last_row(height - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (::abs(y) - (y < 0)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::abs(last_row - ::abs(last_row - y) + (y > last_row)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_high(::abs(y) - (y < 0));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
+        }
+
+        int last_row;
+    };
+
+    template <typename D> struct BrdReflect
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdReflect(int height, int width) : last_row(height - 1), last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdReflect(int height, int width, U) : last_row(height - 1), last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (::abs(y) - (y < 0)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return /*::abs*/(last_row - ::abs(last_row - y) + (y > last_row)) /*% (last_row + 1)*/;
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (::abs(x) - (x < 0)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return (last_col - ::abs(last_col - x) + (x > last_col));
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int last_row;
+        int last_col;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdWrap
+
+    template <typename D> struct BrdRowWrap
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowWrap(int width_) : width(width_) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowWrap(int width_, U) : width(width_) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (x >= 0) * x + (x < 0) * (x - ((x - width + 1) / width) * width);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return (x < width) * x + (x >= width) * (x % width);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_high(idx_col_low(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int width;
+    };
+
+    template <typename D> struct BrdColWrap
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColWrap(int height_) : height(height_) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColWrap(int height_, U) : height(height_) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (y >= 0) * y + (y < 0) * (y - ((y - height + 1) / height) * height);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return (y < height) * y + (y >= height) * (y % height);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_high(idx_row_low(y));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
+        }
+
+        int height;
+    };
+
+    template <typename D> struct BrdWrap
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdWrap(int height_, int width_) :
+            height(height_), width(width_)
+        {
+        }
+        template <typename U>
+        __host__ __device__ __forceinline__ BrdWrap(int height_, int width_, U) :
+            height(height_), width(width_)
+        {
+        }
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (y >= 0) ? y : (y - ((y - height + 1) / height) * height);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return (y < height) ? y : (y % height);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_high(idx_row_low(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (x >= 0) ? x : (x - ((x - width + 1) / width) * width);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return (x < width) ? x : (x % width);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_high(idx_col_low(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int height;
+        int width;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BorderReader
+
+    template <typename Ptr2D, typename B> struct BorderReader
+    {
+        typedef typename B::result_type elem_type;
+        typedef typename Ptr2D::index_type index_type;
+
+        __host__ __device__ __forceinline__ BorderReader(const Ptr2D& ptr_, const B& b_) : ptr(ptr_), b(b_) {}
+
+        __device__ __forceinline__ elem_type operator ()(index_type y, index_type x) const
+        {
+            return b.at(y, x, ptr);
+        }
+
+        Ptr2D ptr;
+        B b;
+    };
+
+    // under win32 there is some bug with templated types that passed as kernel parameters
+    // with this specialization all works fine
+    template <typename Ptr2D, typename D> struct BorderReader< Ptr2D, BrdConstant<D> >
+    {
+        typedef typename BrdConstant<D>::result_type elem_type;
+        typedef typename Ptr2D::index_type index_type;
+
+        __host__ __device__ __forceinline__ BorderReader(const Ptr2D& src_, const BrdConstant<D>& b) :
+            src(src_), height(b.height), width(b.width), val(b.val)
+        {
+        }
+
+        __device__ __forceinline__ D operator ()(index_type y, index_type x) const
+        {
+            return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(src(y, x)) : val;
+        }
+
+        Ptr2D src;
+        int height;
+        int width;
+        D val;
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_BORDER_INTERPOLATE_HPP

3rdPart/include/opencv2/core/cuda/color.hpp

@@ -0,0 +1,309 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_COLOR_HPP
+#define OPENCV_CUDA_COLOR_HPP
+
+#include "detail/color_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    // All OPENCV_CUDA_IMPLEMENT_*_TRAITS(ColorSpace1_to_ColorSpace2, ...) macros implements
+    // template <typename T> class ColorSpace1_to_ColorSpace2_traits
+    // {
+    //     typedef ... functor_type;
+    //     static __host__ __device__ functor_type create_functor();
+    // };
+
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgr_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgr_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgr_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgra_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgra_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgra_to_rgba, 4, 4, 2)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgr_to_bgr555, 3, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgr_to_bgr565, 3, 0, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgb_to_bgr555, 3, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgb_to_bgr565, 3, 2, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgra_to_bgr555, 4, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgra_to_bgr565, 4, 0, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgba_to_bgr555, 4, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgba_to_bgr565, 4, 2, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_rgb, 3, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_rgb, 3, 2, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_bgr, 3, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_bgr, 3, 0, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_rgba, 4, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_rgba, 4, 2, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_bgra, 4, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_bgra, 4, 0, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB_TRAITS(gray_to_bgr, 3)
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB_TRAITS(gray_to_bgra, 4)
+
+    #undef OPENCV_CUDA_IMPLEMENT_GRAY2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB5x5_TRAITS(gray_to_bgr555, 5)
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB5x5_TRAITS(gray_to_bgr565, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_GRAY2RGB5x5_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB5x52GRAY_TRAITS(bgr555_to_gray, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52GRAY_TRAITS(bgr565_to_gray, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB5x52GRAY_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(rgb_to_gray, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(bgr_to_gray, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(rgba_to_gray, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(bgra_to_gray, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgb_to_YCrCb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgba_to_YCrCb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgb_to_YCrCb4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgba_to_YCrCb4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgr_to_YCrCb, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgra_to_YCrCb, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgr_to_YCrCb4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgra_to_YCrCb4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgb_to_xyz, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgba_to_xyz, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgb_to_xyz4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgba_to_xyz4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgr_to_xyz, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgra_to_xyz, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgr_to_xyz4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgra_to_xyz4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgb_to_hsv, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgba_to_hsv, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgb_to_hsv4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgba_to_hsv4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgr_to_hsv, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgra_to_hsv, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgr_to_hsv4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgra_to_hsv4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgb_to_hls, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgba_to_hls, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgb_to_hls4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgba_to_hls4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgr_to_hls, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgra_to_hls, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgr_to_hls4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgra_to_hls4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgb_to_lab, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgba_to_lab, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgb_to_lab4, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgba_to_lab4, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgr_to_lab, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgra_to_lab, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgr_to_lab4, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgra_to_lab4, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgb_to_lab, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgba_to_lab, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgb_to_lab4, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgba_to_lab4, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgr_to_lab, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgra_to_lab, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgr_to_lab4, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgra_to_lab4, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_rgb, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_rgb, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_rgba, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_rgba, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_bgr, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_bgr, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_bgra, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_bgra, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lrgb, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lrgb, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lrgba, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lrgba, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lbgr, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lbgr, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lbgra, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lbgra, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgb_to_luv, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgba_to_luv, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgb_to_luv4, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgba_to_luv4, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgr_to_luv, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgra_to_luv, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgr_to_luv4, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgra_to_luv4, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgb_to_luv, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgba_to_luv, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgb_to_luv4, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgba_to_luv4, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgr_to_luv, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgra_to_luv, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgr_to_luv4, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgra_to_luv4, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_rgb, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_rgb, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_rgba, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_rgba, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_bgr, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_bgr, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_bgra, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_bgra, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lrgb, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lrgb, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lrgba, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lrgba, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lbgr, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lbgr, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lbgra, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lbgra, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_COLOR_HPP

3rdPart/include/opencv2/core/cuda/common.hpp

@@ -0,0 +1,123 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_COMMON_HPP
+#define OPENCV_CUDA_COMMON_HPP
+
+#include <cuda_runtime.h>
+#include "opencv2/core/cuda_types.hpp"
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/base.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+#ifndef CV_PI_F
+    #ifndef CV_PI
+        #define CV_PI_F 3.14159265f
+    #else
+        #define CV_PI_F ((float)CV_PI)
+    #endif
+#endif
+
+namespace cv { namespace cuda {
+    static inline void checkCudaError(cudaError_t err, const char* file, const int line, const char* func)
+    {
+        if (cudaSuccess != err)
+            cv::error(cv::Error::GpuApiCallError, cudaGetErrorString(err), func, file, line);
+    }
+}}
+
+#ifndef cudaSafeCall
+    #define cudaSafeCall(expr)  cv::cuda::checkCudaError(expr, __FILE__, __LINE__, CV_Func)
+#endif
+
+namespace cv { namespace cuda
+{
+    template <typename T> static inline bool isAligned(const T* ptr, size_t size)
+    {
+        return reinterpret_cast<size_t>(ptr) % size == 0;
+    }
+
+    static inline bool isAligned(size_t step, size_t size)
+    {
+        return step % size == 0;
+    }
+}}
+
+namespace cv { namespace cuda
+{
+    namespace device
+    {
+        __host__ __device__ __forceinline__ int divUp(int total, int grain)
+        {
+            return (total + grain - 1) / grain;
+        }
+
+        template<class T> inline void bindTexture(const textureReference* tex, const PtrStepSz<T>& img)
+        {
+            cudaChannelFormatDesc desc = cudaCreateChannelDesc<T>();
+            cudaSafeCall( cudaBindTexture2D(0, tex, img.ptr(), &desc, img.cols, img.rows, img.step) );
+        }
+
+        template<class T> inline void createTextureObjectPitch2D(cudaTextureObject_t* tex, PtrStepSz<T>& img, const cudaTextureDesc& texDesc)
+        {
+            cudaResourceDesc resDesc;
+            memset(&resDesc, 0, sizeof(resDesc));
+            resDesc.resType = cudaResourceTypePitch2D;
+            resDesc.res.pitch2D.devPtr = static_cast<void*>(img.ptr());
+            resDesc.res.pitch2D.height = img.rows;
+            resDesc.res.pitch2D.width = img.cols;
+            resDesc.res.pitch2D.pitchInBytes = img.step;
+            resDesc.res.pitch2D.desc = cudaCreateChannelDesc<T>();
+
+            cudaSafeCall( cudaCreateTextureObject(tex, &resDesc, &texDesc, NULL) );
+        }
+    }
+}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_COMMON_HPP

3rdPart/include/opencv2/core/cuda/datamov_utils.hpp

@@ -0,0 +1,113 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DATAMOV_UTILS_HPP
+#define OPENCV_CUDA_DATAMOV_UTILS_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 200
+
+        // for Fermi memory space is detected automatically
+        template <typename T> struct ForceGlob
+        {
+            __device__ __forceinline__ static void Load(const T* ptr, int offset, T& val)  { val = ptr[offset];  }
+        };
+
+    #else // __CUDA_ARCH__ >= 200
+
+        #if defined(_WIN64) || defined(__LP64__)
+            // 64-bit register modifier for inlined asm
+            #define OPENCV_CUDA_ASM_PTR "l"
+        #else
+            // 32-bit register modifier for inlined asm
+            #define OPENCV_CUDA_ASM_PTR "r"
+        #endif
+
+        template<class T> struct ForceGlob;
+
+        #define OPENCV_CUDA_DEFINE_FORCE_GLOB(base_type, ptx_type, reg_mod) \
+            template <> struct ForceGlob<base_type> \
+            { \
+                __device__ __forceinline__ static void Load(const base_type* ptr, int offset, base_type& val) \
+                { \
+                    asm("ld.global."#ptx_type" %0, [%1];" : "="#reg_mod(val) : OPENCV_CUDA_ASM_PTR(ptr + offset)); \
+                } \
+            };
+
+        #define OPENCV_CUDA_DEFINE_FORCE_GLOB_B(base_type, ptx_type) \
+            template <> struct ForceGlob<base_type> \
+            { \
+                __device__ __forceinline__ static void Load(const base_type* ptr, int offset, base_type& val) \
+                { \
+                    asm("ld.global."#ptx_type" %0, [%1];" : "=r"(*reinterpret_cast<uint*>(&val)) : OPENCV_CUDA_ASM_PTR(ptr + offset)); \
+                } \
+            };
+
+            OPENCV_CUDA_DEFINE_FORCE_GLOB_B(uchar,  u8)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB_B(schar,  s8)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB_B(char,   b8)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (ushort, u16, h)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (short,  s16, h)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (uint,   u32, r)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (int,    s32, r)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (float,  f32, f)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (double, f64, d)
+
+        #undef OPENCV_CUDA_DEFINE_FORCE_GLOB
+        #undef OPENCV_CUDA_DEFINE_FORCE_GLOB_B
+        #undef OPENCV_CUDA_ASM_PTR
+
+    #endif // __CUDA_ARCH__ >= 200
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_DATAMOV_UTILS_HPP

3rdPart/include/opencv2/core/cuda/detail/reduce.hpp

@@ -0,0 +1,365 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_REDUCE_DETAIL_HPP
+#define OPENCV_CUDA_REDUCE_DETAIL_HPP
+
+#include <thrust/tuple.h>
+#include "../warp.hpp"
+#include "../warp_shuffle.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace reduce_detail
+    {
+        template <typename T> struct GetType;
+        template <typename T> struct GetType<T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<volatile T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<T&>
+        {
+            typedef T type;
+        };
+
+        template <unsigned int I, unsigned int N>
+        struct For
+        {
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadToSmem(const PointerTuple& smem, const ValTuple& val, unsigned int tid)
+            {
+                thrust::get<I>(smem)[tid] = thrust::get<I>(val);
+
+                For<I + 1, N>::loadToSmem(smem, val, tid);
+            }
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadFromSmem(const PointerTuple& smem, const ValTuple& val, unsigned int tid)
+            {
+                thrust::get<I>(val) = thrust::get<I>(smem)[tid];
+
+                For<I + 1, N>::loadFromSmem(smem, val, tid);
+            }
+
+            template <class PointerTuple, class ValTuple, class OpTuple>
+            static __device__ void merge(const PointerTuple& smem, const ValTuple& val, unsigned int tid, unsigned int delta, const OpTuple& op)
+            {
+                typename GetType<typename thrust::tuple_element<I, PointerTuple>::type>::type reg = thrust::get<I>(smem)[tid + delta];
+                thrust::get<I>(smem)[tid] = thrust::get<I>(val) = thrust::get<I>(op)(thrust::get<I>(val), reg);
+
+                For<I + 1, N>::merge(smem, val, tid, delta, op);
+            }
+            template <class ValTuple, class OpTuple>
+            static __device__ void mergeShfl(const ValTuple& val, unsigned int delta, unsigned int width, const OpTuple& op)
+            {
+                typename GetType<typename thrust::tuple_element<I, ValTuple>::type>::type reg = shfl_down(thrust::get<I>(val), delta, width);
+                thrust::get<I>(val) = thrust::get<I>(op)(thrust::get<I>(val), reg);
+
+                For<I + 1, N>::mergeShfl(val, delta, width, op);
+            }
+        };
+        template <unsigned int N>
+        struct For<N, N>
+        {
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadToSmem(const PointerTuple&, const ValTuple&, unsigned int)
+            {
+            }
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadFromSmem(const PointerTuple&, const ValTuple&, unsigned int)
+            {
+            }
+
+            template <class PointerTuple, class ValTuple, class OpTuple>
+            static __device__ void merge(const PointerTuple&, const ValTuple&, unsigned int, unsigned int, const OpTuple&)
+            {
+            }
+            template <class ValTuple, class OpTuple>
+            static __device__ void mergeShfl(const ValTuple&, unsigned int, unsigned int, const OpTuple&)
+            {
+            }
+        };
+
+        template <typename T>
+        __device__ __forceinline__ void loadToSmem(volatile T* smem, T& val, unsigned int tid)
+        {
+            smem[tid] = val;
+        }
+        template <typename T>
+        __device__ __forceinline__ void loadFromSmem(volatile T* smem, T& val, unsigned int tid)
+        {
+            val = smem[tid];
+        }
+        template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+                  typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9>
+        __device__ __forceinline__ void loadToSmem(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                                       const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                                       unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::loadToSmem(smem, val, tid);
+        }
+        template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+                  typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9>
+        __device__ __forceinline__ void loadFromSmem(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                                         const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                                         unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::loadFromSmem(smem, val, tid);
+        }
+
+        template <typename T, class Op>
+        __device__ __forceinline__ void merge(volatile T* smem, T& val, unsigned int tid, unsigned int delta, const Op& op)
+        {
+            T reg = smem[tid + delta];
+            smem[tid] = val = op(val, reg);
+        }
+        template <typename T, class Op>
+        __device__ __forceinline__ void mergeShfl(T& val, unsigned int delta, unsigned int width, const Op& op)
+        {
+            T reg = shfl_down(val, delta, width);
+            val = op(val, reg);
+        }
+        template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+                  typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+                  class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+        __device__ __forceinline__ void merge(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                              const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                              unsigned int tid,
+                                              unsigned int delta,
+                                              const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::merge(smem, val, tid, delta, op);
+        }
+        template <typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+                  class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+        __device__ __forceinline__ void mergeShfl(const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                                  unsigned int delta,
+                                                  unsigned int width,
+                                                  const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9> >::value>::mergeShfl(val, delta, width, op);
+        }
+
+        template <unsigned int N> struct Generic
+        {
+            template <typename Pointer, typename Reference, class Op>
+            static __device__ void reduce(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+                loadToSmem(smem, val, tid);
+                if (N >= 32)
+                    __syncthreads();
+
+                if (N >= 2048)
+                {
+                    if (tid < 1024)
+                        merge(smem, val, tid, 1024, op);
+
+                    __syncthreads();
+                }
+                if (N >= 1024)
+                {
+                    if (tid < 512)
+                        merge(smem, val, tid, 512, op);
+
+                    __syncthreads();
+                }
+                if (N >= 512)
+                {
+                    if (tid < 256)
+                        merge(smem, val, tid, 256, op);
+
+                    __syncthreads();
+                }
+                if (N >= 256)
+                {
+                    if (tid < 128)
+                        merge(smem, val, tid, 128, op);
+
+                    __syncthreads();
+                }
+                if (N >= 128)
+                {
+                    if (tid < 64)
+                        merge(smem, val, tid, 64, op);
+
+                    __syncthreads();
+                }
+                if (N >= 64)
+                {
+                    if (tid < 32)
+                        merge(smem, val, tid, 32, op);
+                }
+
+                if (tid < 16)
+                {
+                    merge(smem, val, tid, 16, op);
+                    merge(smem, val, tid, 8, op);
+                    merge(smem, val, tid, 4, op);
+                    merge(smem, val, tid, 2, op);
+                    merge(smem, val, tid, 1, op);
+                }
+            }
+        };
+
+        template <unsigned int I, typename Pointer, typename Reference, class Op>
+        struct Unroll
+        {
+            static __device__ void loopShfl(Reference val, Op op, unsigned int N)
+            {
+                mergeShfl(val, I, N, op);
+                Unroll<I / 2, Pointer, Reference, Op>::loopShfl(val, op, N);
+            }
+            static __device__ void loop(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+                merge(smem, val, tid, I, op);
+                Unroll<I / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+            }
+        };
+        template <typename Pointer, typename Reference, class Op>
+        struct Unroll<0, Pointer, Reference, Op>
+        {
+            static __device__ void loopShfl(Reference, Op, unsigned int)
+            {
+            }
+            static __device__ void loop(Pointer, Reference, unsigned int, Op)
+            {
+            }
+        };
+
+        template <unsigned int N> struct WarpOptimized
+        {
+            template <typename Pointer, typename Reference, class Op>
+            static __device__ void reduce(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+            #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+                CV_UNUSED(smem);
+                CV_UNUSED(tid);
+
+                Unroll<N / 2, Pointer, Reference, Op>::loopShfl(val, op, N);
+            #else
+                loadToSmem(smem, val, tid);
+
+                if (tid < N / 2)
+                    Unroll<N / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+            #endif
+            }
+        };
+
+        template <unsigned int N> struct GenericOptimized32
+        {
+            enum { M = N / 32 };
+
+            template <typename Pointer, typename Reference, class Op>
+            static __device__ void reduce(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+                const unsigned int laneId = Warp::laneId();
+
+            #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+                Unroll<16, Pointer, Reference, Op>::loopShfl(val, op, warpSize);
+
+                if (laneId == 0)
+                    loadToSmem(smem, val, tid / 32);
+            #else
+                loadToSmem(smem, val, tid);
+
+                if (laneId < 16)
+                    Unroll<16, Pointer, Reference, Op>::loop(smem, val, tid, op);
+
+                __syncthreads();
+
+                if (laneId == 0)
+                    loadToSmem(smem, val, tid / 32);
+            #endif
+
+                __syncthreads();
+
+                loadFromSmem(smem, val, tid);
+
+                if (tid < 32)
+                {
+                #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+                    Unroll<M / 2, Pointer, Reference, Op>::loopShfl(val, op, M);
+                #else
+                    Unroll<M / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+                #endif
+                }
+            }
+        };
+
+        template <bool val, class T1, class T2> struct StaticIf;
+        template <class T1, class T2> struct StaticIf<true, T1, T2>
+        {
+            typedef T1 type;
+        };
+        template <class T1, class T2> struct StaticIf<false, T1, T2>
+        {
+            typedef T2 type;
+        };
+
+        template <unsigned int N> struct IsPowerOf2
+        {
+            enum { value = ((N != 0) && !(N & (N - 1))) };
+        };
+
+        template <unsigned int N> struct Dispatcher
+        {
+            typedef typename StaticIf<
+                (N <= 32) && IsPowerOf2<N>::value,
+                WarpOptimized<N>,
+                typename StaticIf<
+                    (N <= 1024) && IsPowerOf2<N>::value,
+                    GenericOptimized32<N>,
+                    Generic<N>
+                >::type
+            >::type reductor;
+        };
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_REDUCE_DETAIL_HPP

3rdPart/include/opencv2/core/cuda/detail/reduce_key_val.hpp

@@ -0,0 +1,502 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_PRED_VAL_REDUCE_DETAIL_HPP
+#define OPENCV_CUDA_PRED_VAL_REDUCE_DETAIL_HPP
+
+#include <thrust/tuple.h>
+#include "../warp.hpp"
+#include "../warp_shuffle.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace reduce_key_val_detail
+    {
+        template <typename T> struct GetType;
+        template <typename T> struct GetType<T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<volatile T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<T&>
+        {
+            typedef T type;
+        };
+
+        template <unsigned int I, unsigned int N>
+        struct For
+        {
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadToSmem(const PointerTuple& smem, const ReferenceTuple& data, unsigned int tid)
+            {
+                thrust::get<I>(smem)[tid] = thrust::get<I>(data);
+
+                For<I + 1, N>::loadToSmem(smem, data, tid);
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadFromSmem(const PointerTuple& smem, const ReferenceTuple& data, unsigned int tid)
+            {
+                thrust::get<I>(data) = thrust::get<I>(smem)[tid];
+
+                For<I + 1, N>::loadFromSmem(smem, data, tid);
+            }
+
+            template <class ReferenceTuple>
+            static __device__ void copyShfl(const ReferenceTuple& val, unsigned int delta, int width)
+            {
+                thrust::get<I>(val) = shfl_down(thrust::get<I>(val), delta, width);
+
+                For<I + 1, N>::copyShfl(val, delta, width);
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void copy(const PointerTuple& svals, const ReferenceTuple& val, unsigned int tid, unsigned int delta)
+            {
+                thrust::get<I>(svals)[tid] = thrust::get<I>(val) = thrust::get<I>(svals)[tid + delta];
+
+                For<I + 1, N>::copy(svals, val, tid, delta);
+            }
+
+            template <class KeyReferenceTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void mergeShfl(const KeyReferenceTuple& key, const ValReferenceTuple& val, const CmpTuple& cmp, unsigned int delta, int width)
+            {
+                typename GetType<typename thrust::tuple_element<I, KeyReferenceTuple>::type>::type reg = shfl_down(thrust::get<I>(key), delta, width);
+
+                if (thrust::get<I>(cmp)(reg, thrust::get<I>(key)))
+                {
+                    thrust::get<I>(key) = reg;
+                    thrust::get<I>(val) = shfl_down(thrust::get<I>(val), delta, width);
+                }
+
+                For<I + 1, N>::mergeShfl(key, val, cmp, delta, width);
+            }
+            template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void merge(const KeyPointerTuple& skeys, const KeyReferenceTuple& key,
+                                         const ValPointerTuple& svals, const ValReferenceTuple& val,
+                                         const CmpTuple& cmp,
+                                         unsigned int tid, unsigned int delta)
+            {
+                typename GetType<typename thrust::tuple_element<I, KeyPointerTuple>::type>::type reg = thrust::get<I>(skeys)[tid + delta];
+
+                if (thrust::get<I>(cmp)(reg, thrust::get<I>(key)))
+                {
+                    thrust::get<I>(skeys)[tid] = thrust::get<I>(key) = reg;
+                    thrust::get<I>(svals)[tid] = thrust::get<I>(val) = thrust::get<I>(svals)[tid + delta];
+                }
+
+                For<I + 1, N>::merge(skeys, key, svals, val, cmp, tid, delta);
+            }
+        };
+        template <unsigned int N>
+        struct For<N, N>
+        {
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadToSmem(const PointerTuple&, const ReferenceTuple&, unsigned int)
+            {
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadFromSmem(const PointerTuple&, const ReferenceTuple&, unsigned int)
+            {
+            }
+
+            template <class ReferenceTuple>
+            static __device__ void copyShfl(const ReferenceTuple&, unsigned int, int)
+            {
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void copy(const PointerTuple&, const ReferenceTuple&, unsigned int, unsigned int)
+            {
+            }
+
+            template <class KeyReferenceTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void mergeShfl(const KeyReferenceTuple&, const ValReferenceTuple&, const CmpTuple&, unsigned int, int)
+            {
+            }
+            template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void merge(const KeyPointerTuple&, const KeyReferenceTuple&,
+                                         const ValPointerTuple&, const ValReferenceTuple&,
+                                         const CmpTuple&,
+                                         unsigned int, unsigned int)
+            {
+            }
+        };
+
+        //////////////////////////////////////////////////////
+        // loadToSmem
+
+        template <typename T>
+        __device__ __forceinline__ void loadToSmem(volatile T* smem, T& data, unsigned int tid)
+        {
+            smem[tid] = data;
+        }
+        template <typename T>
+        __device__ __forceinline__ void loadFromSmem(volatile T* smem, T& data, unsigned int tid)
+        {
+            data = smem[tid];
+        }
+        template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void loadToSmem(const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& smem,
+                                                   const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& data,
+                                                   unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::loadToSmem(smem, data, tid);
+        }
+        template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void loadFromSmem(const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& smem,
+                                                     const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& data,
+                                                     unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::loadFromSmem(smem, data, tid);
+        }
+
+        //////////////////////////////////////////////////////
+        // copyVals
+
+        template <typename V>
+        __device__ __forceinline__ void copyValsShfl(V& val, unsigned int delta, int width)
+        {
+            val = shfl_down(val, delta, width);
+        }
+        template <typename V>
+        __device__ __forceinline__ void copyVals(volatile V* svals, V& val, unsigned int tid, unsigned int delta)
+        {
+            svals[tid] = val = svals[tid + delta];
+        }
+        template <typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void copyValsShfl(const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                     unsigned int delta,
+                                                     int width)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9> >::value>::copyShfl(val, delta, width);
+        }
+        template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void copyVals(const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 unsigned int tid, unsigned int delta)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::copy(svals, val, tid, delta);
+        }
+
+        //////////////////////////////////////////////////////
+        // merge
+
+        template <typename K, typename V, class Cmp>
+        __device__ __forceinline__ void mergeShfl(K& key, V& val, const Cmp& cmp, unsigned int delta, int width)
+        {
+            K reg = shfl_down(key, delta, width);
+
+            if (cmp(reg, key))
+            {
+                key = reg;
+                copyValsShfl(val, delta, width);
+            }
+        }
+        template <typename K, typename V, class Cmp>
+        __device__ __forceinline__ void merge(volatile K* skeys, K& key, volatile V* svals, V& val, const Cmp& cmp, unsigned int tid, unsigned int delta)
+        {
+            K reg = skeys[tid + delta];
+
+            if (cmp(reg, key))
+            {
+                skeys[tid] = key = reg;
+                copyVals(svals, val, tid, delta);
+            }
+        }
+        template <typename K,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp>
+        __device__ __forceinline__ void mergeShfl(K& key,
+                                                  const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                  const Cmp& cmp,
+                                                  unsigned int delta, int width)
+        {
+            K reg = shfl_down(key, delta, width);
+
+            if (cmp(reg, key))
+            {
+                key = reg;
+                copyValsShfl(val, delta, width);
+            }
+        }
+        template <typename K,
+                  typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp>
+        __device__ __forceinline__ void merge(volatile K* skeys, K& key,
+                                              const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                              const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                              const Cmp& cmp, unsigned int tid, unsigned int delta)
+        {
+            K reg = skeys[tid + delta];
+
+            if (cmp(reg, key))
+            {
+                skeys[tid] = key = reg;
+                copyVals(svals, val, tid, delta);
+            }
+        }
+        template <typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+        __device__ __forceinline__ void mergeShfl(const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                                  const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                  const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp,
+                                                  unsigned int delta, int width)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9> >::value>::mergeShfl(key, val, cmp, delta, width);
+        }
+        template <typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+                  typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+                  typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+        __device__ __forceinline__ void merge(const thrust::tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                              const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                              const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                              const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                              const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp,
+                                              unsigned int tid, unsigned int delta)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::merge(skeys, key, svals, val, cmp, tid, delta);
+        }
+
+        //////////////////////////////////////////////////////
+        // Generic
+
+        template <unsigned int N> struct Generic
+        {
+            template <class KP, class KR, class VP, class VR, class Cmp>
+            static __device__ void reduce(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+                loadToSmem(skeys, key, tid);
+                loadValsToSmem(svals, val, tid);
+                if (N >= 32)
+                    __syncthreads();
+
+                if (N >= 2048)
+                {
+                    if (tid < 1024)
+                        merge(skeys, key, svals, val, cmp, tid, 1024);
+
+                    __syncthreads();
+                }
+                if (N >= 1024)
+                {
+                    if (tid < 512)
+                        merge(skeys, key, svals, val, cmp, tid, 512);
+
+                    __syncthreads();
+                }
+                if (N >= 512)
+                {
+                    if (tid < 256)
+                        merge(skeys, key, svals, val, cmp, tid, 256);
+
+                    __syncthreads();
+                }
+                if (N >= 256)
+                {
+                    if (tid < 128)
+                        merge(skeys, key, svals, val, cmp, tid, 128);
+
+                    __syncthreads();
+                }
+                if (N >= 128)
+                {
+                    if (tid < 64)
+                        merge(skeys, key, svals, val, cmp, tid, 64);
+
+                    __syncthreads();
+                }
+                if (N >= 64)
+                {
+                    if (tid < 32)
+                        merge(skeys, key, svals, val, cmp, tid, 32);
+                }
+
+                if (tid < 16)
+                {
+                    merge(skeys, key, svals, val, cmp, tid, 16);
+                    merge(skeys, key, svals, val, cmp, tid, 8);
+                    merge(skeys, key, svals, val, cmp, tid, 4);
+                    merge(skeys, key, svals, val, cmp, tid, 2);
+                    merge(skeys, key, svals, val, cmp, tid, 1);
+                }
+            }
+        };
+
+        template <unsigned int I, class KP, class KR, class VP, class VR, class Cmp>
+        struct Unroll
+        {
+            static __device__ void loopShfl(KR key, VR val, Cmp cmp, unsigned int N)
+            {
+                mergeShfl(key, val, cmp, I, N);
+                Unroll<I / 2, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, N);
+            }
+            static __device__ void loop(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+                merge(skeys, key, svals, val, cmp, tid, I);
+                Unroll<I / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+            }
+        };
+        template <class KP, class KR, class VP, class VR, class Cmp>
+        struct Unroll<0, KP, KR, VP, VR, Cmp>
+        {
+            static __device__ void loopShfl(KR, VR, Cmp, unsigned int)
+            {
+            }
+            static __device__ void loop(KP, KR, VP, VR, unsigned int, Cmp)
+            {
+            }
+        };
+
+        template <unsigned int N> struct WarpOptimized
+        {
+            template <class KP, class KR, class VP, class VR, class Cmp>
+            static __device__ void reduce(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+            #if 0 // __CUDA_ARCH__ >= 300
+                CV_UNUSED(skeys);
+                CV_UNUSED(svals);
+                CV_UNUSED(tid);
+
+                Unroll<N / 2, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, N);
+            #else
+                loadToSmem(skeys, key, tid);
+                loadToSmem(svals, val, tid);
+
+                if (tid < N / 2)
+                    Unroll<N / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+            #endif
+            }
+        };
+
+        template <unsigned int N> struct GenericOptimized32
+        {
+            enum { M = N / 32 };
+
+            template <class KP, class KR, class VP, class VR, class Cmp>
+            static __device__ void reduce(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+                const unsigned int laneId = Warp::laneId();
+
+            #if 0 // __CUDA_ARCH__ >= 300
+                Unroll<16, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, warpSize);
+
+                if (laneId == 0)
+                {
+                    loadToSmem(skeys, key, tid / 32);
+                    loadToSmem(svals, val, tid / 32);
+                }
+            #else
+                loadToSmem(skeys, key, tid);
+                loadToSmem(svals, val, tid);
+
+                if (laneId < 16)
+                    Unroll<16, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+
+                __syncthreads();
+
+                if (laneId == 0)
+                {
+                    loadToSmem(skeys, key, tid / 32);
+                    loadToSmem(svals, val, tid / 32);
+                }
+            #endif
+
+                __syncthreads();
+
+                loadFromSmem(skeys, key, tid);
+
+                if (tid < 32)
+                {
+                #if 0 // __CUDA_ARCH__ >= 300
+                    loadFromSmem(svals, val, tid);
+
+                    Unroll<M / 2, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, M);
+                #else
+                    Unroll<M / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+                #endif
+                }
+            }
+        };
+
+        template <bool val, class T1, class T2> struct StaticIf;
+        template <class T1, class T2> struct StaticIf<true, T1, T2>
+        {
+            typedef T1 type;
+        };
+        template <class T1, class T2> struct StaticIf<false, T1, T2>
+        {
+            typedef T2 type;
+        };
+
+        template <unsigned int N> struct IsPowerOf2
+        {
+            enum { value = ((N != 0) && !(N & (N - 1))) };
+        };
+
+        template <unsigned int N> struct Dispatcher
+        {
+            typedef typename StaticIf<
+                (N <= 32) && IsPowerOf2<N>::value,
+                WarpOptimized<N>,
+                typename StaticIf<
+                    (N <= 1024) && IsPowerOf2<N>::value,
+                    GenericOptimized32<N>,
+                    Generic<N>
+                >::type
+            >::type reductor;
+        };
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_PRED_VAL_REDUCE_DETAIL_HPP

3rdPart/include/opencv2/core/cuda/detail/transform_detail.hpp

@@ -0,0 +1,392 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TRANSFORM_DETAIL_HPP
+#define OPENCV_CUDA_TRANSFORM_DETAIL_HPP
+
+#include "../common.hpp"
+#include "../vec_traits.hpp"
+#include "../functional.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace transform_detail
+    {
+        //! Read Write Traits
+
+        template <typename T, typename D, int shift> struct UnaryReadWriteTraits
+        {
+            typedef typename TypeVec<T, shift>::vec_type read_type;
+            typedef typename TypeVec<D, shift>::vec_type write_type;
+        };
+
+        template <typename T1, typename T2, typename D, int shift> struct BinaryReadWriteTraits
+        {
+            typedef typename TypeVec<T1, shift>::vec_type read_type1;
+            typedef typename TypeVec<T2, shift>::vec_type read_type2;
+            typedef typename TypeVec<D, shift>::vec_type write_type;
+        };
+
+        //! Transform kernels
+
+        template <int shift> struct OpUnroller;
+        template <> struct OpUnroller<1>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+            }
+        };
+        template <> struct OpUnroller<2>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src.y);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src1.y, src2.y);
+            }
+        };
+        template <> struct OpUnroller<3>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, const UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src.z);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, const BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src1.y, src2.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src1.z, src2.z);
+            }
+        };
+        template <> struct OpUnroller<4>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, const UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src.z);
+                if (mask(y, x_shifted + 3))
+                    dst.w = op(src.w);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, const BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src1.y, src2.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src1.z, src2.z);
+                if (mask(y, x_shifted + 3))
+                    dst.w = op(src1.w, src2.w);
+            }
+        };
+        template <> struct OpUnroller<8>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, const UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.a0 = op(src.a0);
+                if (mask(y, x_shifted + 1))
+                    dst.a1 = op(src.a1);
+                if (mask(y, x_shifted + 2))
+                    dst.a2 = op(src.a2);
+                if (mask(y, x_shifted + 3))
+                    dst.a3 = op(src.a3);
+                if (mask(y, x_shifted + 4))
+                    dst.a4 = op(src.a4);
+                if (mask(y, x_shifted + 5))
+                    dst.a5 = op(src.a5);
+                if (mask(y, x_shifted + 6))
+                    dst.a6 = op(src.a6);
+                if (mask(y, x_shifted + 7))
+                    dst.a7 = op(src.a7);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, const BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.a0 = op(src1.a0, src2.a0);
+                if (mask(y, x_shifted + 1))
+                    dst.a1 = op(src1.a1, src2.a1);
+                if (mask(y, x_shifted + 2))
+                    dst.a2 = op(src1.a2, src2.a2);
+                if (mask(y, x_shifted + 3))
+                    dst.a3 = op(src1.a3, src2.a3);
+                if (mask(y, x_shifted + 4))
+                    dst.a4 = op(src1.a4, src2.a4);
+                if (mask(y, x_shifted + 5))
+                    dst.a5 = op(src1.a5, src2.a5);
+                if (mask(y, x_shifted + 6))
+                    dst.a6 = op(src1.a6, src2.a6);
+                if (mask(y, x_shifted + 7))
+                    dst.a7 = op(src1.a7, src2.a7);
+            }
+        };
+
+        template <typename T, typename D, typename UnOp, typename Mask>
+        static __global__ void transformSmart(const PtrStepSz<T> src_, PtrStep<D> dst_, const Mask mask, const UnOp op)
+        {
+            typedef TransformFunctorTraits<UnOp> ft;
+            typedef typename UnaryReadWriteTraits<T, D, ft::smart_shift>::read_type read_type;
+            typedef typename UnaryReadWriteTraits<T, D, ft::smart_shift>::write_type write_type;
+
+            const int x = threadIdx.x + blockIdx.x * blockDim.x;
+            const int y = threadIdx.y + blockIdx.y * blockDim.y;
+            const int x_shifted = x * ft::smart_shift;
+
+            if (y < src_.rows)
+            {
+                const T* src = src_.ptr(y);
+                D* dst = dst_.ptr(y);
+
+                if (x_shifted + ft::smart_shift - 1 < src_.cols)
+                {
+                    const read_type src_n_el = ((const read_type*)src)[x];
+                    OpUnroller<ft::smart_shift>::unroll(src_n_el, ((write_type*)dst)[x], mask, op, x_shifted, y);
+                }
+                else
+                {
+                    for (int real_x = x_shifted; real_x < src_.cols; ++real_x)
+                    {
+                        if (mask(y, real_x))
+                            dst[real_x] = op(src[real_x]);
+                    }
+                }
+            }
+        }
+
+        template <typename T, typename D, typename UnOp, typename Mask>
+        __global__ static void transformSimple(const PtrStepSz<T> src, PtrStep<D> dst, const Mask mask, const UnOp op)
+        {
+            const int x = blockDim.x * blockIdx.x + threadIdx.x;
+            const int y = blockDim.y * blockIdx.y + threadIdx.y;
+
+            if (x < src.cols && y < src.rows && mask(y, x))
+            {
+                dst.ptr(y)[x] = op(src.ptr(y)[x]);
+            }
+        }
+
+        template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+        static __global__ void transformSmart(const PtrStepSz<T1> src1_, const PtrStep<T2> src2_, PtrStep<D> dst_,
+            const Mask mask, const BinOp op)
+        {
+            typedef TransformFunctorTraits<BinOp> ft;
+            typedef typename BinaryReadWriteTraits<T1, T2, D, ft::smart_shift>::read_type1 read_type1;
+            typedef typename BinaryReadWriteTraits<T1, T2, D, ft::smart_shift>::read_type2 read_type2;
+            typedef typename BinaryReadWriteTraits<T1, T2, D, ft::smart_shift>::write_type write_type;
+
+            const int x = threadIdx.x + blockIdx.x * blockDim.x;
+            const int y = threadIdx.y + blockIdx.y * blockDim.y;
+            const int x_shifted = x * ft::smart_shift;
+
+            if (y < src1_.rows)
+            {
+                const T1* src1 = src1_.ptr(y);
+                const T2* src2 = src2_.ptr(y);
+                D* dst = dst_.ptr(y);
+
+                if (x_shifted + ft::smart_shift - 1 < src1_.cols)
+                {
+                    const read_type1 src1_n_el = ((const read_type1*)src1)[x];
+                    const read_type2 src2_n_el = ((const read_type2*)src2)[x];
+
+                    OpUnroller<ft::smart_shift>::unroll(src1_n_el, src2_n_el, ((write_type*)dst)[x], mask, op, x_shifted, y);
+                }
+                else
+                {
+                    for (int real_x = x_shifted; real_x < src1_.cols; ++real_x)
+                    {
+                        if (mask(y, real_x))
+                            dst[real_x] = op(src1[real_x], src2[real_x]);
+                    }
+                }
+            }
+        }
+
+        template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+        static __global__ void transformSimple(const PtrStepSz<T1> src1, const PtrStep<T2> src2, PtrStep<D> dst,
+            const Mask mask, const BinOp op)
+        {
+            const int x = blockDim.x * blockIdx.x + threadIdx.x;
+            const int y = blockDim.y * blockIdx.y + threadIdx.y;
+
+            if (x < src1.cols && y < src1.rows && mask(y, x))
+            {
+                const T1 src1_data = src1.ptr(y)[x];
+                const T2 src2_data = src2.ptr(y)[x];
+                dst.ptr(y)[x] = op(src1_data, src2_data);
+            }
+        }
+
+        template <bool UseSmart> struct TransformDispatcher;
+        template<> struct TransformDispatcher<false>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static void call(PtrStepSz<T> src, PtrStepSz<D> dst, UnOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<UnOp> ft;
+
+                const dim3 threads(ft::simple_block_dim_x, ft::simple_block_dim_y, 1);
+                const dim3 grid(divUp(src.cols, threads.x), divUp(src.rows, threads.y), 1);
+
+                transformSimple<T, D><<<grid, threads, 0, stream>>>(src, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static void call(PtrStepSz<T1> src1, PtrStepSz<T2> src2, PtrStepSz<D> dst, BinOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<BinOp> ft;
+
+                const dim3 threads(ft::simple_block_dim_x, ft::simple_block_dim_y, 1);
+                const dim3 grid(divUp(src1.cols, threads.x), divUp(src1.rows, threads.y), 1);
+
+                transformSimple<T1, T2, D><<<grid, threads, 0, stream>>>(src1, src2, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+        };
+        template<> struct TransformDispatcher<true>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static void call(PtrStepSz<T> src, PtrStepSz<D> dst, UnOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<UnOp> ft;
+
+                CV_StaticAssert(ft::smart_shift != 1, "");
+
+                if (!isAligned(src.data, ft::smart_shift * sizeof(T)) || !isAligned(src.step, ft::smart_shift * sizeof(T)) ||
+                    !isAligned(dst.data, ft::smart_shift * sizeof(D)) || !isAligned(dst.step, ft::smart_shift * sizeof(D)))
+                {
+                    TransformDispatcher<false>::call(src, dst, op, mask, stream);
+                    return;
+                }
+
+                const dim3 threads(ft::smart_block_dim_x, ft::smart_block_dim_y, 1);
+                const dim3 grid(divUp(src.cols, threads.x * ft::smart_shift), divUp(src.rows, threads.y), 1);
+
+                transformSmart<T, D><<<grid, threads, 0, stream>>>(src, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static void call(PtrStepSz<T1> src1, PtrStepSz<T2> src2, PtrStepSz<D> dst, BinOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<BinOp> ft;
+
+                CV_StaticAssert(ft::smart_shift != 1, "");
+
+                if (!isAligned(src1.data, ft::smart_shift * sizeof(T1)) || !isAligned(src1.step, ft::smart_shift * sizeof(T1)) ||
+                    !isAligned(src2.data, ft::smart_shift * sizeof(T2)) || !isAligned(src2.step, ft::smart_shift * sizeof(T2)) ||
+                    !isAligned(dst.data, ft::smart_shift * sizeof(D)) || !isAligned(dst.step, ft::smart_shift * sizeof(D)))
+                {
+                    TransformDispatcher<false>::call(src1, src2, dst, op, mask, stream);
+                    return;
+                }
+
+                const dim3 threads(ft::smart_block_dim_x, ft::smart_block_dim_y, 1);
+                const dim3 grid(divUp(src1.cols, threads.x * ft::smart_shift), divUp(src1.rows, threads.y), 1);
+
+                transformSmart<T1, T2, D><<<grid, threads, 0, stream>>>(src1, src2, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+        };
+    } // namespace transform_detail
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TRANSFORM_DETAIL_HPP

3rdPart/include/opencv2/core/cuda/detail/type_traits_detail.hpp

@@ -0,0 +1,191 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TYPE_TRAITS_DETAIL_HPP
+#define OPENCV_CUDA_TYPE_TRAITS_DETAIL_HPP
+
+#include "../common.hpp"
+#include "../vec_traits.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace type_traits_detail
+    {
+        template <bool, typename T1, typename T2> struct Select { typedef T1 type; };
+        template <typename T1, typename T2> struct Select<false, T1, T2> { typedef T2 type; };
+
+        template <typename T> struct IsSignedIntergral { enum {value = 0}; };
+        template <> struct IsSignedIntergral<schar> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<char1> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<short> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<short1> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<int> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<int1> { enum {value = 1}; };
+
+        template <typename T> struct IsUnsignedIntegral { enum {value = 0}; };
+        template <> struct IsUnsignedIntegral<uchar> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<uchar1> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<ushort> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<ushort1> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<uint> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<uint1> { enum {value = 1}; };
+
+        template <typename T> struct IsIntegral { enum {value = IsSignedIntergral<T>::value || IsUnsignedIntegral<T>::value}; };
+        template <> struct IsIntegral<char> { enum {value = 1}; };
+        template <> struct IsIntegral<bool> { enum {value = 1}; };
+
+        template <typename T> struct IsFloat { enum {value = 0}; };
+        template <> struct IsFloat<float> { enum {value = 1}; };
+        template <> struct IsFloat<double> { enum {value = 1}; };
+
+        template <typename T> struct IsVec { enum {value = 0}; };
+        template <> struct IsVec<uchar1> { enum {value = 1}; };
+        template <> struct IsVec<uchar2> { enum {value = 1}; };
+        template <> struct IsVec<uchar3> { enum {value = 1}; };
+        template <> struct IsVec<uchar4> { enum {value = 1}; };
+        template <> struct IsVec<uchar8> { enum {value = 1}; };
+        template <> struct IsVec<char1> { enum {value = 1}; };
+        template <> struct IsVec<char2> { enum {value = 1}; };
+        template <> struct IsVec<char3> { enum {value = 1}; };
+        template <> struct IsVec<char4> { enum {value = 1}; };
+        template <> struct IsVec<char8> { enum {value = 1}; };
+        template <> struct IsVec<ushort1> { enum {value = 1}; };
+        template <> struct IsVec<ushort2> { enum {value = 1}; };
+        template <> struct IsVec<ushort3> { enum {value = 1}; };
+        template <> struct IsVec<ushort4> { enum {value = 1}; };
+        template <> struct IsVec<ushort8> { enum {value = 1}; };
+        template <> struct IsVec<short1> { enum {value = 1}; };
+        template <> struct IsVec<short2> { enum {value = 1}; };
+        template <> struct IsVec<short3> { enum {value = 1}; };
+        template <> struct IsVec<short4> { enum {value = 1}; };
+        template <> struct IsVec<short8> { enum {value = 1}; };
+        template <> struct IsVec<uint1> { enum {value = 1}; };
+        template <> struct IsVec<uint2> { enum {value = 1}; };
+        template <> struct IsVec<uint3> { enum {value = 1}; };
+        template <> struct IsVec<uint4> { enum {value = 1}; };
+        template <> struct IsVec<uint8> { enum {value = 1}; };
+        template <> struct IsVec<int1> { enum {value = 1}; };
+        template <> struct IsVec<int2> { enum {value = 1}; };
+        template <> struct IsVec<int3> { enum {value = 1}; };
+        template <> struct IsVec<int4> { enum {value = 1}; };
+        template <> struct IsVec<int8> { enum {value = 1}; };
+        template <> struct IsVec<float1> { enum {value = 1}; };
+        template <> struct IsVec<float2> { enum {value = 1}; };
+        template <> struct IsVec<float3> { enum {value = 1}; };
+        template <> struct IsVec<float4> { enum {value = 1}; };
+        template <> struct IsVec<float8> { enum {value = 1}; };
+        template <> struct IsVec<double1> { enum {value = 1}; };
+        template <> struct IsVec<double2> { enum {value = 1}; };
+        template <> struct IsVec<double3> { enum {value = 1}; };
+        template <> struct IsVec<double4> { enum {value = 1}; };
+        template <> struct IsVec<double8> { enum {value = 1}; };
+
+        template <class U> struct AddParameterType { typedef const U& type; };
+        template <class U> struct AddParameterType<U&> { typedef U& type; };
+        template <> struct AddParameterType<void> { typedef void type; };
+
+        template <class U> struct ReferenceTraits
+        {
+            enum { value = false };
+            typedef U type;
+        };
+        template <class U> struct ReferenceTraits<U&>
+        {
+            enum { value = true };
+            typedef U type;
+        };
+
+        template <class U> struct PointerTraits
+        {
+            enum { value = false };
+            typedef void type;
+        };
+        template <class U> struct PointerTraits<U*>
+        {
+            enum { value = true };
+            typedef U type;
+        };
+        template <class U> struct PointerTraits<U*&>
+        {
+            enum { value = true };
+            typedef U type;
+        };
+
+        template <class U> struct UnConst
+        {
+            typedef U type;
+            enum { value = 0 };
+        };
+        template <class U> struct UnConst<const U>
+        {
+            typedef U type;
+            enum { value = 1 };
+        };
+        template <class U> struct UnConst<const U&>
+        {
+            typedef U& type;
+            enum { value = 1 };
+        };
+
+        template <class U> struct UnVolatile
+        {
+            typedef U type;
+            enum { value = 0 };
+        };
+        template <class U> struct UnVolatile<volatile U>
+        {
+            typedef U type;
+            enum { value = 1 };
+        };
+        template <class U> struct UnVolatile<volatile U&>
+        {
+            typedef U& type;
+            enum { value = 1 };
+        };
+    } // namespace type_traits_detail
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TYPE_TRAITS_DETAIL_HPP

3rdPart/include/opencv2/core/cuda/detail/vec_distance_detail.hpp

@@ -0,0 +1,121 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VEC_DISTANCE_DETAIL_HPP
+#define OPENCV_CUDA_VEC_DISTANCE_DETAIL_HPP
+
+#include "../datamov_utils.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace vec_distance_detail
+    {
+        template <int THREAD_DIM, int N> struct UnrollVecDiffCached
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ void calcCheck(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, int ind)
+            {
+                if (ind < len)
+                {
+                    T1 val1 = *vecCached++;
+
+                    T2 val2;
+                    ForceGlob<T2>::Load(vecGlob, ind, val2);
+
+                    dist.reduceIter(val1, val2);
+
+                    UnrollVecDiffCached<THREAD_DIM, N - 1>::calcCheck(vecCached, vecGlob, len, dist, ind + THREAD_DIM);
+                }
+            }
+
+            template <typename Dist, typename T1, typename T2>
+            static __device__ void calcWithoutCheck(const T1* vecCached, const T2* vecGlob, Dist& dist)
+            {
+                T1 val1 = *vecCached++;
+
+                T2 val2;
+                ForceGlob<T2>::Load(vecGlob, 0, val2);
+                vecGlob += THREAD_DIM;
+
+                dist.reduceIter(val1, val2);
+
+                UnrollVecDiffCached<THREAD_DIM, N - 1>::calcWithoutCheck(vecCached, vecGlob, dist);
+            }
+        };
+        template <int THREAD_DIM> struct UnrollVecDiffCached<THREAD_DIM, 0>
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calcCheck(const T1*, const T2*, int, Dist&, int)
+            {
+            }
+
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calcWithoutCheck(const T1*, const T2*, Dist&)
+            {
+            }
+        };
+
+        template <int THREAD_DIM, int MAX_LEN, bool LEN_EQ_MAX_LEN> struct VecDiffCachedCalculator;
+        template <int THREAD_DIM, int MAX_LEN> struct VecDiffCachedCalculator<THREAD_DIM, MAX_LEN, false>
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calc(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, int tid)
+            {
+                UnrollVecDiffCached<THREAD_DIM, MAX_LEN / THREAD_DIM>::calcCheck(vecCached, vecGlob, len, dist, tid);
+            }
+        };
+        template <int THREAD_DIM, int MAX_LEN> struct VecDiffCachedCalculator<THREAD_DIM, MAX_LEN, true>
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calc(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, int tid)
+            {
+                UnrollVecDiffCached<THREAD_DIM, MAX_LEN / THREAD_DIM>::calcWithoutCheck(vecCached, vecGlob + tid, dist);
+            }
+        };
+    } // namespace vec_distance_detail
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VEC_DISTANCE_DETAIL_HPP

3rdPart/include/opencv2/core/cuda/dynamic_smem.hpp

@@ -0,0 +1,88 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DYNAMIC_SMEM_HPP
+#define OPENCV_CUDA_DYNAMIC_SMEM_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<class T> struct DynamicSharedMem
+    {
+        __device__ __forceinline__ operator T*()
+        {
+            extern __shared__ int __smem[];
+            return (T*)__smem;
+        }
+
+        __device__ __forceinline__ operator const T*() const
+        {
+            extern __shared__ int __smem[];
+            return (T*)__smem;
+        }
+    };
+
+    // specialize for double to avoid unaligned memory access compile errors
+    template<> struct DynamicSharedMem<double>
+    {
+        __device__ __forceinline__ operator double*()
+        {
+            extern __shared__ double __smem_d[];
+            return (double*)__smem_d;
+        }
+
+        __device__ __forceinline__ operator const double*() const
+        {
+            extern __shared__ double __smem_d[];
+            return (double*)__smem_d;
+        }
+    };
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_DYNAMIC_SMEM_HPP

3rdPart/include/opencv2/core/cuda/emulation.hpp

@@ -0,0 +1,269 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_EMULATION_HPP_
+#define OPENCV_CUDA_EMULATION_HPP_
+
+#include "common.hpp"
+#include "warp_reduce.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct Emulation
+    {
+
+        static __device__ __forceinline__ int syncthreadsOr(int pred)
+        {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 200)
+                // just campilation stab
+                return 0;
+#else
+                return __syncthreads_or(pred);
+#endif
+        }
+
+        template<int CTA_SIZE>
+        static __forceinline__ __device__ int Ballot(int predicate)
+        {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ >= 200)
+            return __ballot(predicate);
+#else
+            __shared__ volatile int cta_buffer[CTA_SIZE];
+
+            int tid = threadIdx.x;
+            cta_buffer[tid] = predicate ? (1 << (tid & 31)) : 0;
+            return warp_reduce(cta_buffer);
+#endif
+        }
+
+        struct smem
+        {
+            enum { TAG_MASK = (1U << ( (sizeof(unsigned int) << 3) - 5U)) - 1U };
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicInc(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count;
+                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
+                do
+                {
+                    count = *address & TAG_MASK;
+                    count = tag | (count + 1);
+                    *address = count;
+                } while (*address != count);
+
+                return (count & TAG_MASK) - 1;
+#else
+                return ::atomicInc(address, val);
+#endif
+            }
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicAdd(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count;
+                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
+                do
+                {
+                    count = *address & TAG_MASK;
+                    count = tag | (count + val);
+                    *address = count;
+                } while (*address != count);
+
+                return (count & TAG_MASK) - val;
+#else
+                return ::atomicAdd(address, val);
+#endif
+            }
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicMin(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count = ::min(*address, val);
+                do
+                {
+                    *address = count;
+                } while (*address > count);
+
+                return count;
+#else
+                return ::atomicMin(address, val);
+#endif
+            }
+        }; // struct cmem
+
+        struct glob
+        {
+            static __device__ __forceinline__ int atomicAdd(int* address, int val)
+            {
+                return ::atomicAdd(address, val);
+            }
+            static __device__ __forceinline__ unsigned int atomicAdd(unsigned int* address, unsigned int val)
+            {
+                return ::atomicAdd(address, val);
+            }
+            static __device__ __forceinline__ float atomicAdd(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 200
+                return ::atomicAdd(address, val);
+            #else
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(val + __int_as_float(assumed)));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #endif
+            }
+            static __device__ __forceinline__ double atomicAdd(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(val + __longlong_as_double(assumed)));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0;
+            #endif
+            }
+
+            static __device__ __forceinline__ int atomicMin(int* address, int val)
+            {
+                return ::atomicMin(address, val);
+            }
+            static __device__ __forceinline__ float atomicMin(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 120
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(::fminf(val, __int_as_float(assumed))));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0f;
+            #endif
+            }
+            static __device__ __forceinline__ double atomicMin(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(::fmin(val, __longlong_as_double(assumed))));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0;
+            #endif
+            }
+
+            static __device__ __forceinline__ int atomicMax(int* address, int val)
+            {
+                return ::atomicMax(address, val);
+            }
+            static __device__ __forceinline__ float atomicMax(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 120
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(::fmaxf(val, __int_as_float(assumed))));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0f;
+            #endif
+            }
+            static __device__ __forceinline__ double atomicMax(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(::fmax(val, __longlong_as_double(assumed))));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0;
+            #endif
+            }
+        };
+    }; //struct Emulation
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_EMULATION_HPP_ */

3rdPart/include/opencv2/core/cuda/filters.hpp

@@ -0,0 +1,286 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_FILTERS_HPP
+#define OPENCV_CUDA_FILTERS_HPP
+
+#include "saturate_cast.hpp"
+#include "vec_traits.hpp"
+#include "vec_math.hpp"
+#include "type_traits.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename Ptr2D> struct PointFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ PointFilter(const Ptr2D& src_, float fx = 0.f, float fy = 0.f)
+        : src(src_)
+        {
+            CV_UNUSED(fx);
+            CV_UNUSED(fy);
+        }
+
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            return src(__float2int_rz(y), __float2int_rz(x));
+        }
+
+        Ptr2D src;
+    };
+
+    template <typename Ptr2D> struct LinearFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ LinearFilter(const Ptr2D& src_, float fx = 0.f, float fy = 0.f)
+        : src(src_)
+        {
+            CV_UNUSED(fx);
+            CV_UNUSED(fy);
+        }
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+
+            work_type out = VecTraits<work_type>::all(0);
+
+            const int x1 = __float2int_rd(x);
+            const int y1 = __float2int_rd(y);
+            const int x2 = x1 + 1;
+            const int y2 = y1 + 1;
+
+            elem_type src_reg = src(y1, x1);
+            out = out + src_reg * ((x2 - x) * (y2 - y));
+
+            src_reg = src(y1, x2);
+            out = out + src_reg * ((x - x1) * (y2 - y));
+
+            src_reg = src(y2, x1);
+            out = out + src_reg * ((x2 - x) * (y - y1));
+
+            src_reg = src(y2, x2);
+            out = out + src_reg * ((x - x1) * (y - y1));
+
+            return saturate_cast<elem_type>(out);
+        }
+
+        Ptr2D src;
+    };
+
+    template <typename Ptr2D> struct CubicFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+        typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+
+        explicit __host__ __device__ __forceinline__ CubicFilter(const Ptr2D& src_, float fx = 0.f, float fy = 0.f)
+        : src(src_)
+        {
+            CV_UNUSED(fx);
+            CV_UNUSED(fy);
+        }
+
+        static __device__ __forceinline__ float bicubicCoeff(float x_)
+        {
+            float x = fabsf(x_);
+            if (x <= 1.0f)
+            {
+                return x * x * (1.5f * x - 2.5f) + 1.0f;
+            }
+            else if (x < 2.0f)
+            {
+                return x * (x * (-0.5f * x + 2.5f) - 4.0f) + 2.0f;
+            }
+            else
+            {
+                return 0.0f;
+            }
+        }
+
+        __device__ elem_type operator ()(float y, float x) const
+        {
+            const float xmin = ::ceilf(x - 2.0f);
+            const float xmax = ::floorf(x + 2.0f);
+
+            const float ymin = ::ceilf(y - 2.0f);
+            const float ymax = ::floorf(y + 2.0f);
+
+            work_type sum = VecTraits<work_type>::all(0);
+            float wsum = 0.0f;
+
+            for (float cy = ymin; cy <= ymax; cy += 1.0f)
+            {
+                for (float cx = xmin; cx <= xmax; cx += 1.0f)
+                {
+                    const float w = bicubicCoeff(x - cx) * bicubicCoeff(y - cy);
+                    sum = sum + w * src(__float2int_rd(cy), __float2int_rd(cx));
+                    wsum += w;
+                }
+            }
+
+            work_type res = (!wsum)? VecTraits<work_type>::all(0) : sum / wsum;
+
+            return saturate_cast<elem_type>(res);
+        }
+
+        Ptr2D src;
+    };
+    // for integer scaling
+    template <typename Ptr2D> struct IntegerAreaFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ IntegerAreaFilter(const Ptr2D& src_, float scale_x_, float scale_y_)
+            : src(src_), scale_x(scale_x_), scale_y(scale_y_), scale(1.f / (scale_x * scale_y)) {}
+
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            float fsx1 = x * scale_x;
+            float fsx2 = fsx1 + scale_x;
+
+            int sx1 = __float2int_ru(fsx1);
+            int sx2 = __float2int_rd(fsx2);
+
+            float fsy1 = y * scale_y;
+            float fsy2 = fsy1 + scale_y;
+
+            int sy1 = __float2int_ru(fsy1);
+            int sy2 = __float2int_rd(fsy2);
+
+            typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+            work_type out = VecTraits<work_type>::all(0.f);
+
+            for(int dy = sy1; dy < sy2; ++dy)
+                for(int dx = sx1; dx < sx2; ++dx)
+                {
+                    out = out + src(dy, dx) * scale;
+                }
+
+            return saturate_cast<elem_type>(out);
+        }
+
+        Ptr2D src;
+        float scale_x, scale_y ,scale;
+    };
+
+    template <typename Ptr2D> struct AreaFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ AreaFilter(const Ptr2D& src_, float scale_x_, float scale_y_)
+            : src(src_), scale_x(scale_x_), scale_y(scale_y_){}
+
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            float fsx1 = x * scale_x;
+            float fsx2 = fsx1 + scale_x;
+
+            int sx1 = __float2int_ru(fsx1);
+            int sx2 = __float2int_rd(fsx2);
+
+            float fsy1 = y * scale_y;
+            float fsy2 = fsy1 + scale_y;
+
+            int sy1 = __float2int_ru(fsy1);
+            int sy2 = __float2int_rd(fsy2);
+
+            float scale = 1.f / (fminf(scale_x, src.width - fsx1) * fminf(scale_y, src.height - fsy1));
+
+            typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+            work_type out = VecTraits<work_type>::all(0.f);
+
+            for (int dy = sy1; dy < sy2; ++dy)
+            {
+                for (int dx = sx1; dx < sx2; ++dx)
+                    out = out + src(dy, dx) * scale;
+
+                if (sx1 > fsx1)
+                    out = out + src(dy, (sx1 -1) ) * ((sx1 - fsx1) * scale);
+
+                if (sx2 < fsx2)
+                    out = out + src(dy, sx2) * ((fsx2 -sx2) * scale);
+            }
+
+            if (sy1 > fsy1)
+                for (int dx = sx1; dx < sx2; ++dx)
+                    out = out + src( (sy1 - 1) , dx) * ((sy1 -fsy1) * scale);
+
+            if (sy2 < fsy2)
+                for (int dx = sx1; dx < sx2; ++dx)
+                    out = out + src(sy2, dx) * ((fsy2 -sy2) * scale);
+
+            if ((sy1 > fsy1) &&  (sx1 > fsx1))
+                out = out + src( (sy1 - 1) , (sx1 - 1)) * ((sy1 -fsy1) * (sx1 -fsx1) * scale);
+
+            if ((sy1 > fsy1) &&  (sx2 < fsx2))
+                out = out + src( (sy1 - 1) , sx2) * ((sy1 -fsy1) * (fsx2 -sx2) * scale);
+
+            if ((sy2 < fsy2) &&  (sx2 < fsx2))
+                out = out + src(sy2, sx2) * ((fsy2 -sy2) * (fsx2 -sx2) * scale);
+
+            if ((sy2 < fsy2) &&  (sx1 > fsx1))
+                out = out + src(sy2, (sx1 - 1)) * ((fsy2 -sy2) * (sx1 -fsx1) * scale);
+
+            return saturate_cast<elem_type>(out);
+        }
+
+        Ptr2D src;
+        float scale_x, scale_y;
+        int width, haight;
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_FILTERS_HPP

3rdPart/include/opencv2/core/cuda/funcattrib.hpp

@@ -0,0 +1,79 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DEVICE_FUNCATTRIB_HPP
+#define OPENCV_CUDA_DEVICE_FUNCATTRIB_HPP
+
+#include <cstdio>
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<class Func>
+    void printFuncAttrib(Func& func)
+    {
+
+        cudaFuncAttributes attrs;
+        cudaFuncGetAttributes(&attrs, func);
+
+        printf("=== Function stats ===\n");
+        printf("Name: \n");
+        printf("sharedSizeBytes    = %d\n", attrs.sharedSizeBytes);
+        printf("constSizeBytes     = %d\n", attrs.constSizeBytes);
+        printf("localSizeBytes     = %d\n", attrs.localSizeBytes);
+        printf("maxThreadsPerBlock = %d\n", attrs.maxThreadsPerBlock);
+        printf("numRegs            = %d\n", attrs.numRegs);
+        printf("ptxVersion         = %d\n", attrs.ptxVersion);
+        printf("binaryVersion      = %d\n", attrs.binaryVersion);
+        printf("\n");
+        fflush(stdout);
+    }
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif  /* OPENCV_CUDA_DEVICE_FUNCATTRIB_HPP */

3rdPart/include/opencv2/core/cuda/functional.hpp

@@ -0,0 +1,810 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_FUNCTIONAL_HPP
+#define OPENCV_CUDA_FUNCTIONAL_HPP
+
+#include <functional>
+#include "saturate_cast.hpp"
+#include "vec_traits.hpp"
+#include "type_traits.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    // Function Objects
+#ifdef CV_CXX11
+    template<typename Argument, typename Result> struct unary_function
+    {
+        typedef Argument argument_type;
+        typedef Result result_type;
+    };
+    template<typename Argument1, typename Argument2, typename Result> struct binary_function
+    {
+        typedef Argument1 first_argument_type;
+        typedef Argument2 second_argument_type;
+        typedef Result result_type;
+    };
+#else
+    template<typename Argument, typename Result> struct unary_function : public std::unary_function<Argument, Result> {};
+    template<typename Argument1, typename Argument2, typename Result> struct binary_function : public std::binary_function<Argument1, Argument2, Result> {};
+#endif
+
+    // Arithmetic Operations
+    template <typename T> struct plus : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a + b;
+        }
+        __host__ __device__ __forceinline__ plus() {}
+        __host__ __device__ __forceinline__ plus(const plus&) {}
+    };
+
+    template <typename T> struct minus : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a - b;
+        }
+        __host__ __device__ __forceinline__ minus() {}
+        __host__ __device__ __forceinline__ minus(const minus&) {}
+    };
+
+    template <typename T> struct multiplies : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a * b;
+        }
+        __host__ __device__ __forceinline__ multiplies() {}
+        __host__ __device__ __forceinline__ multiplies(const multiplies&) {}
+    };
+
+    template <typename T> struct divides : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a / b;
+        }
+        __host__ __device__ __forceinline__ divides() {}
+        __host__ __device__ __forceinline__ divides(const divides&) {}
+    };
+
+    template <typename T> struct modulus : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a % b;
+        }
+        __host__ __device__ __forceinline__ modulus() {}
+        __host__ __device__ __forceinline__ modulus(const modulus&) {}
+    };
+
+    template <typename T> struct negate : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a) const
+        {
+            return -a;
+        }
+        __host__ __device__ __forceinline__ negate() {}
+        __host__ __device__ __forceinline__ negate(const negate&) {}
+    };
+
+    // Comparison Operations
+    template <typename T> struct equal_to : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a == b;
+        }
+        __host__ __device__ __forceinline__ equal_to() {}
+        __host__ __device__ __forceinline__ equal_to(const equal_to&) {}
+    };
+
+    template <typename T> struct not_equal_to : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a != b;
+        }
+        __host__ __device__ __forceinline__ not_equal_to() {}
+        __host__ __device__ __forceinline__ not_equal_to(const not_equal_to&) {}
+    };
+
+    template <typename T> struct greater : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a > b;
+        }
+        __host__ __device__ __forceinline__ greater() {}
+        __host__ __device__ __forceinline__ greater(const greater&) {}
+    };
+
+    template <typename T> struct less : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a < b;
+        }
+        __host__ __device__ __forceinline__ less() {}
+        __host__ __device__ __forceinline__ less(const less&) {}
+    };
+
+    template <typename T> struct greater_equal : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a >= b;
+        }
+        __host__ __device__ __forceinline__ greater_equal() {}
+        __host__ __device__ __forceinline__ greater_equal(const greater_equal&) {}
+    };
+
+    template <typename T> struct less_equal : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a <= b;
+        }
+        __host__ __device__ __forceinline__ less_equal() {}
+        __host__ __device__ __forceinline__ less_equal(const less_equal&) {}
+    };
+
+    // Logical Operations
+    template <typename T> struct logical_and : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a && b;
+        }
+        __host__ __device__ __forceinline__ logical_and() {}
+        __host__ __device__ __forceinline__ logical_and(const logical_and&) {}
+    };
+
+    template <typename T> struct logical_or : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a || b;
+        }
+        __host__ __device__ __forceinline__ logical_or() {}
+        __host__ __device__ __forceinline__ logical_or(const logical_or&) {}
+    };
+
+    template <typename T> struct logical_not : unary_function<T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a) const
+        {
+            return !a;
+        }
+        __host__ __device__ __forceinline__ logical_not() {}
+        __host__ __device__ __forceinline__ logical_not(const logical_not&) {}
+    };
+
+    // Bitwise Operations
+    template <typename T> struct bit_and : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a & b;
+        }
+        __host__ __device__ __forceinline__ bit_and() {}
+        __host__ __device__ __forceinline__ bit_and(const bit_and&) {}
+    };
+
+    template <typename T> struct bit_or : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a | b;
+        }
+        __host__ __device__ __forceinline__ bit_or() {}
+        __host__ __device__ __forceinline__ bit_or(const bit_or&) {}
+    };
+
+    template <typename T> struct bit_xor : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a ^ b;
+        }
+        __host__ __device__ __forceinline__ bit_xor() {}
+        __host__ __device__ __forceinline__ bit_xor(const bit_xor&) {}
+    };
+
+    template <typename T> struct bit_not : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType v) const
+        {
+            return ~v;
+        }
+        __host__ __device__ __forceinline__ bit_not() {}
+        __host__ __device__ __forceinline__ bit_not(const bit_not&) {}
+    };
+
+    // Generalized Identity Operations
+    template <typename T> struct identity : unary_function<T, T>
+    {
+        __device__ __forceinline__ typename TypeTraits<T>::ParameterType operator()(typename TypeTraits<T>::ParameterType x) const
+        {
+            return x;
+        }
+        __host__ __device__ __forceinline__ identity() {}
+        __host__ __device__ __forceinline__ identity(const identity&) {}
+    };
+
+    template <typename T1, typename T2> struct project1st : binary_function<T1, T2, T1>
+    {
+        __device__ __forceinline__ typename TypeTraits<T1>::ParameterType operator()(typename TypeTraits<T1>::ParameterType lhs, typename TypeTraits<T2>::ParameterType rhs) const
+        {
+            return lhs;
+        }
+        __host__ __device__ __forceinline__ project1st() {}
+        __host__ __device__ __forceinline__ project1st(const project1st&) {}
+    };
+
+    template <typename T1, typename T2> struct project2nd : binary_function<T1, T2, T2>
+    {
+        __device__ __forceinline__ typename TypeTraits<T2>::ParameterType operator()(typename TypeTraits<T1>::ParameterType lhs, typename TypeTraits<T2>::ParameterType rhs) const
+        {
+            return rhs;
+        }
+        __host__ __device__ __forceinline__ project2nd() {}
+        __host__ __device__ __forceinline__ project2nd(const project2nd&) {}
+    };
+
+    // Min/Max Operations
+
+#define OPENCV_CUDA_IMPLEMENT_MINMAX(name, type, op) \
+    template <> struct name<type> : binary_function<type, type, type> \
+    { \
+        __device__ __forceinline__ type operator()(type lhs, type rhs) const {return op(lhs, rhs);} \
+        __host__ __device__ __forceinline__ name() {}\
+        __host__ __device__ __forceinline__ name(const name&) {}\
+    };
+
+    template <typename T> struct maximum : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType lhs, typename TypeTraits<T>::ParameterType rhs) const
+        {
+            return max(lhs, rhs);
+        }
+        __host__ __device__ __forceinline__ maximum() {}
+        __host__ __device__ __forceinline__ maximum(const maximum&) {}
+    };
+
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, uchar, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, schar, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, char, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, ushort, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, short, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, int, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, uint, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, float, ::fmax)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, double, ::fmax)
+
+    template <typename T> struct minimum : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType lhs, typename TypeTraits<T>::ParameterType rhs) const
+        {
+            return min(lhs, rhs);
+        }
+        __host__ __device__ __forceinline__ minimum() {}
+        __host__ __device__ __forceinline__ minimum(const minimum&) {}
+    };
+
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, uchar, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, schar, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, char, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, ushort, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, short, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, int, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, uint, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, float, ::fmin)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, double, ::fmin)
+
+#undef OPENCV_CUDA_IMPLEMENT_MINMAX
+
+    // Math functions
+
+    template <typename T> struct abs_func : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType x) const
+        {
+            return abs(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<unsigned char> : unary_function<unsigned char, unsigned char>
+    {
+        __device__ __forceinline__ unsigned char operator ()(unsigned char x) const
+        {
+            return x;
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<signed char> : unary_function<signed char, signed char>
+    {
+        __device__ __forceinline__ signed char operator ()(signed char x) const
+        {
+            return ::abs((int)x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<char> : unary_function<char, char>
+    {
+        __device__ __forceinline__ char operator ()(char x) const
+        {
+            return ::abs((int)x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<unsigned short> : unary_function<unsigned short, unsigned short>
+    {
+        __device__ __forceinline__ unsigned short operator ()(unsigned short x) const
+        {
+            return x;
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<short> : unary_function<short, short>
+    {
+        __device__ __forceinline__ short operator ()(short x) const
+        {
+            return ::abs((int)x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<unsigned int> : unary_function<unsigned int, unsigned int>
+    {
+        __device__ __forceinline__ unsigned int operator ()(unsigned int x) const
+        {
+            return x;
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<int> : unary_function<int, int>
+    {
+        __device__ __forceinline__ int operator ()(int x) const
+        {
+            return ::abs(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<float> : unary_function<float, float>
+    {
+        __device__ __forceinline__ float operator ()(float x) const
+        {
+            return ::fabsf(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<double> : unary_function<double, double>
+    {
+        __device__ __forceinline__ double operator ()(double x) const
+        {
+            return ::fabs(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+
+#define OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(name, func) \
+    template <typename T> struct name ## _func : unary_function<T, float> \
+    { \
+        __device__ __forceinline__ float operator ()(typename TypeTraits<T>::ParameterType v) const \
+        { \
+            return func ## f(v); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    }; \
+    template <> struct name ## _func<double> : unary_function<double, double> \
+    { \
+        __device__ __forceinline__ double operator ()(double v) const \
+        { \
+            return func(v); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    };
+
+#define OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(name, func) \
+    template <typename T> struct name ## _func : binary_function<T, T, float> \
+    { \
+        __device__ __forceinline__ float operator ()(typename TypeTraits<T>::ParameterType v1, typename TypeTraits<T>::ParameterType v2) const \
+        { \
+            return func ## f(v1, v2); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    }; \
+    template <> struct name ## _func<double> : binary_function<double, double, double> \
+    { \
+        __device__ __forceinline__ double operator ()(double v1, double v2) const \
+        { \
+            return func(v1, v2); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    };
+
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(sqrt, ::sqrt)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(exp, ::exp)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(exp2, ::exp2)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(exp10, ::exp10)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(log, ::log)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(log2, ::log2)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(log10, ::log10)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(sin, ::sin)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(cos, ::cos)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(tan, ::tan)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(asin, ::asin)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(acos, ::acos)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(atan, ::atan)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(sinh, ::sinh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(cosh, ::cosh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(tanh, ::tanh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(asinh, ::asinh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(acosh, ::acosh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(atanh, ::atanh)
+
+    OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(hypot, ::hypot)
+    OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(atan2, ::atan2)
+    OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(pow, ::pow)
+
+    #undef OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR
+    #undef OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR_NO_DOUBLE
+    #undef OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR
+
+    template<typename T> struct hypot_sqr_func : binary_function<T, T, float>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType src1, typename TypeTraits<T>::ParameterType src2) const
+        {
+            return src1 * src1 + src2 * src2;
+        }
+        __host__ __device__ __forceinline__ hypot_sqr_func() {}
+        __host__ __device__ __forceinline__ hypot_sqr_func(const hypot_sqr_func&) {}
+    };
+
+    // Saturate Cast Functor
+    template <typename T, typename D> struct saturate_cast_func : unary_function<T, D>
+    {
+        __device__ __forceinline__ D operator ()(typename TypeTraits<T>::ParameterType v) const
+        {
+            return saturate_cast<D>(v);
+        }
+        __host__ __device__ __forceinline__ saturate_cast_func() {}
+        __host__ __device__ __forceinline__ saturate_cast_func(const saturate_cast_func&) {}
+    };
+
+    // Threshold Functors
+    template <typename T> struct thresh_binary_func : unary_function<T, T>
+    {
+        __host__ __device__ __forceinline__ thresh_binary_func(T thresh_, T maxVal_) : thresh(thresh_), maxVal(maxVal_) {}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src > thresh) * maxVal;
+        }
+
+        __host__ __device__ __forceinline__ thresh_binary_func() {}
+        __host__ __device__ __forceinline__ thresh_binary_func(const thresh_binary_func& other)
+            : thresh(other.thresh), maxVal(other.maxVal) {}
+
+        T thresh;
+        T maxVal;
+    };
+
+    template <typename T> struct thresh_binary_inv_func : unary_function<T, T>
+    {
+        __host__ __device__ __forceinline__ thresh_binary_inv_func(T thresh_, T maxVal_) : thresh(thresh_), maxVal(maxVal_) {}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src <= thresh) * maxVal;
+        }
+
+        __host__ __device__ __forceinline__ thresh_binary_inv_func() {}
+        __host__ __device__ __forceinline__ thresh_binary_inv_func(const thresh_binary_inv_func& other)
+            : thresh(other.thresh), maxVal(other.maxVal) {}
+
+        T thresh;
+        T maxVal;
+    };
+
+    template <typename T> struct thresh_trunc_func : unary_function<T, T>
+    {
+        explicit __host__ __device__ __forceinline__ thresh_trunc_func(T thresh_, T maxVal_ = 0) : thresh(thresh_) {CV_UNUSED(maxVal_);}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return minimum<T>()(src, thresh);
+        }
+
+        __host__ __device__ __forceinline__ thresh_trunc_func() {}
+        __host__ __device__ __forceinline__ thresh_trunc_func(const thresh_trunc_func& other)
+            : thresh(other.thresh) {}
+
+        T thresh;
+    };
+
+    template <typename T> struct thresh_to_zero_func : unary_function<T, T>
+    {
+        explicit __host__ __device__ __forceinline__ thresh_to_zero_func(T thresh_, T maxVal_ = 0) : thresh(thresh_) {CV_UNUSED(maxVal_);}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src > thresh) * src;
+        }
+
+        __host__ __device__ __forceinline__ thresh_to_zero_func() {}
+       __host__  __device__ __forceinline__ thresh_to_zero_func(const thresh_to_zero_func& other)
+            : thresh(other.thresh) {}
+
+        T thresh;
+    };
+
+    template <typename T> struct thresh_to_zero_inv_func : unary_function<T, T>
+    {
+        explicit __host__ __device__ __forceinline__ thresh_to_zero_inv_func(T thresh_, T maxVal_ = 0) : thresh(thresh_) {CV_UNUSED(maxVal_);}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src <= thresh) * src;
+        }
+
+        __host__ __device__ __forceinline__ thresh_to_zero_inv_func() {}
+        __host__ __device__ __forceinline__ thresh_to_zero_inv_func(const thresh_to_zero_inv_func& other)
+            : thresh(other.thresh) {}
+
+        T thresh;
+    };
+
+    // Function Object Adaptors
+    template <typename Predicate> struct unary_negate : unary_function<typename Predicate::argument_type, bool>
+    {
+      explicit __host__ __device__ __forceinline__ unary_negate(const Predicate& p) : pred(p) {}
+
+      __device__ __forceinline__ bool operator()(typename TypeTraits<typename Predicate::argument_type>::ParameterType x) const
+      {
+          return !pred(x);
+      }
+
+      __host__ __device__ __forceinline__ unary_negate() {}
+      __host__ __device__ __forceinline__ unary_negate(const unary_negate& other) : pred(other.pred) {}
+
+      Predicate pred;
+    };
+
+    template <typename Predicate> __host__ __device__ __forceinline__ unary_negate<Predicate> not1(const Predicate& pred)
+    {
+        return unary_negate<Predicate>(pred);
+    }
+
+    template <typename Predicate> struct binary_negate : binary_function<typename Predicate::first_argument_type, typename Predicate::second_argument_type, bool>
+    {
+        explicit __host__ __device__ __forceinline__ binary_negate(const Predicate& p) : pred(p) {}
+
+        __device__ __forceinline__ bool operator()(typename TypeTraits<typename Predicate::first_argument_type>::ParameterType x,
+                                                   typename TypeTraits<typename Predicate::second_argument_type>::ParameterType y) const
+        {
+            return !pred(x,y);
+        }
+
+        __host__ __device__ __forceinline__ binary_negate() {}
+        __host__ __device__ __forceinline__ binary_negate(const binary_negate& other) : pred(other.pred) {}
+
+        Predicate pred;
+    };
+
+    template <typename BinaryPredicate> __host__ __device__ __forceinline__ binary_negate<BinaryPredicate> not2(const BinaryPredicate& pred)
+    {
+        return binary_negate<BinaryPredicate>(pred);
+    }
+
+    template <typename Op> struct binder1st : unary_function<typename Op::second_argument_type, typename Op::result_type>
+    {
+        __host__ __device__ __forceinline__ binder1st(const Op& op_, const typename Op::first_argument_type& arg1_) : op(op_), arg1(arg1_) {}
+
+        __device__ __forceinline__ typename Op::result_type operator ()(typename TypeTraits<typename Op::second_argument_type>::ParameterType a) const
+        {
+            return op(arg1, a);
+        }
+
+        __host__ __device__ __forceinline__ binder1st() {}
+        __host__ __device__ __forceinline__ binder1st(const binder1st& other) : op(other.op), arg1(other.arg1) {}
+
+        Op op;
+        typename Op::first_argument_type arg1;
+    };
+
+    template <typename Op, typename T> __host__ __device__ __forceinline__ binder1st<Op> bind1st(const Op& op, const T& x)
+    {
+        return binder1st<Op>(op, typename Op::first_argument_type(x));
+    }
+
+    template <typename Op> struct binder2nd : unary_function<typename Op::first_argument_type, typename Op::result_type>
+    {
+        __host__ __device__ __forceinline__ binder2nd(const Op& op_, const typename Op::second_argument_type& arg2_) : op(op_), arg2(arg2_) {}
+
+        __forceinline__ __device__ typename Op::result_type operator ()(typename TypeTraits<typename Op::first_argument_type>::ParameterType a) const
+        {
+            return op(a, arg2);
+        }
+
+        __host__ __device__ __forceinline__ binder2nd() {}
+        __host__ __device__ __forceinline__ binder2nd(const binder2nd& other) : op(other.op), arg2(other.arg2) {}
+
+        Op op;
+        typename Op::second_argument_type arg2;
+    };
+
+    template <typename Op, typename T> __host__ __device__ __forceinline__ binder2nd<Op> bind2nd(const Op& op, const T& x)
+    {
+        return binder2nd<Op>(op, typename Op::second_argument_type(x));
+    }
+
+    // Functor Traits
+    template <typename F> struct IsUnaryFunction
+    {
+        typedef char Yes;
+        struct No {Yes a[2];};
+
+        template <typename T, typename D> static Yes check(unary_function<T, D>);
+        static No check(...);
+
+        static F makeF();
+
+        enum { value = (sizeof(check(makeF())) == sizeof(Yes)) };
+    };
+
+    template <typename F> struct IsBinaryFunction
+    {
+        typedef char Yes;
+        struct No {Yes a[2];};
+
+        template <typename T1, typename T2, typename D> static Yes check(binary_function<T1, T2, D>);
+        static No check(...);
+
+        static F makeF();
+
+        enum { value = (sizeof(check(makeF())) == sizeof(Yes)) };
+    };
+
+    namespace functional_detail
+    {
+        template <size_t src_elem_size, size_t dst_elem_size> struct UnOpShift { enum { shift = 1 }; };
+        template <size_t src_elem_size> struct UnOpShift<src_elem_size, 1> { enum { shift = 4 }; };
+        template <size_t src_elem_size> struct UnOpShift<src_elem_size, 2> { enum { shift = 2 }; };
+
+        template <typename T, typename D> struct DefaultUnaryShift
+        {
+            enum { shift = UnOpShift<sizeof(T), sizeof(D)>::shift };
+        };
+
+        template <size_t src_elem_size1, size_t src_elem_size2, size_t dst_elem_size> struct BinOpShift { enum { shift = 1 }; };
+        template <size_t src_elem_size1, size_t src_elem_size2> struct BinOpShift<src_elem_size1, src_elem_size2, 1> { enum { shift = 4 }; };
+        template <size_t src_elem_size1, size_t src_elem_size2> struct BinOpShift<src_elem_size1, src_elem_size2, 2> { enum { shift = 2 }; };
+
+        template <typename T1, typename T2, typename D> struct DefaultBinaryShift
+        {
+            enum { shift = BinOpShift<sizeof(T1), sizeof(T2), sizeof(D)>::shift };
+        };
+
+        template <typename Func, bool unary = IsUnaryFunction<Func>::value> struct ShiftDispatcher;
+        template <typename Func> struct ShiftDispatcher<Func, true>
+        {
+            enum { shift = DefaultUnaryShift<typename Func::argument_type, typename Func::result_type>::shift };
+        };
+        template <typename Func> struct ShiftDispatcher<Func, false>
+        {
+            enum { shift = DefaultBinaryShift<typename Func::first_argument_type, typename Func::second_argument_type, typename Func::result_type>::shift };
+        };
+    }
+
+    template <typename Func> struct DefaultTransformShift
+    {
+        enum { shift = functional_detail::ShiftDispatcher<Func>::shift };
+    };
+
+    template <typename Func> struct DefaultTransformFunctorTraits
+    {
+        enum { simple_block_dim_x = 16 };
+        enum { simple_block_dim_y = 16 };
+
+        enum { smart_block_dim_x = 16 };
+        enum { smart_block_dim_y = 16 };
+        enum { smart_shift = DefaultTransformShift<Func>::shift };
+    };
+
+    template <typename Func> struct TransformFunctorTraits : DefaultTransformFunctorTraits<Func> {};
+
+#define OPENCV_CUDA_TRANSFORM_FUNCTOR_TRAITS(type) \
+    template <> struct TransformFunctorTraits< type > : DefaultTransformFunctorTraits< type >
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_FUNCTIONAL_HPP

3rdPart/include/opencv2/core/cuda/limits.hpp

@@ -0,0 +1,128 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_LIMITS_HPP
+#define OPENCV_CUDA_LIMITS_HPP
+
+#include <limits.h>
+#include <float.h>
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+template <class T> struct numeric_limits;
+
+template <> struct numeric_limits<bool>
+{
+    __device__ __forceinline__ static bool min() { return false; }
+    __device__ __forceinline__ static bool max() { return true;  }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<signed char>
+{
+    __device__ __forceinline__ static signed char min() { return SCHAR_MIN; }
+    __device__ __forceinline__ static signed char max() { return SCHAR_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned char>
+{
+    __device__ __forceinline__ static unsigned char min() { return 0; }
+    __device__ __forceinline__ static unsigned char max() { return UCHAR_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<short>
+{
+    __device__ __forceinline__ static short min() { return SHRT_MIN; }
+    __device__ __forceinline__ static short max() { return SHRT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned short>
+{
+    __device__ __forceinline__ static unsigned short min() { return 0; }
+    __device__ __forceinline__ static unsigned short max() { return USHRT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<int>
+{
+    __device__ __forceinline__ static int min() { return INT_MIN; }
+    __device__ __forceinline__ static int max() { return INT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned int>
+{
+    __device__ __forceinline__ static unsigned int min() { return 0; }
+    __device__ __forceinline__ static unsigned int max() { return UINT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<float>
+{
+    __device__ __forceinline__ static float min() { return FLT_MIN; }
+    __device__ __forceinline__ static float max() { return FLT_MAX; }
+    __device__ __forceinline__ static float epsilon() { return FLT_EPSILON; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<double>
+{
+    __device__ __forceinline__ static double min() { return DBL_MIN; }
+    __device__ __forceinline__ static double max() { return DBL_MAX; }
+    __device__ __forceinline__ static double epsilon() { return DBL_EPSILON; }
+    static const bool is_signed = true;
+};
+}}} // namespace cv { namespace cuda { namespace cudev {
+
+//! @endcond
+
+#endif // OPENCV_CUDA_LIMITS_HPP

3rdPart/include/opencv2/core/cuda/reduce.hpp

@@ -0,0 +1,209 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_REDUCE_HPP
+#define OPENCV_CUDA_REDUCE_HPP
+
+#ifndef THRUST_DEBUG // eliminate -Wundef warning
+#define THRUST_DEBUG 0
+#endif
+
+#include <thrust/tuple.h>
+#include "detail/reduce.hpp"
+#include "detail/reduce_key_val.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <int N, typename T, class Op>
+    __device__ __forceinline__ void reduce(volatile T* smem, T& val, unsigned int tid, const Op& op)
+    {
+        reduce_detail::Dispatcher<N>::reductor::template reduce<volatile T*, T&, const Op&>(smem, val, tid, op);
+    }
+    template <int N,
+              typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+              typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+              class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+    __device__ __forceinline__ void reduce(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                           const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                           unsigned int tid,
+                                           const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+    {
+        reduce_detail::Dispatcher<N>::reductor::template reduce<
+                const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>&,
+                const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>&,
+                const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>&>(smem, val, tid, op);
+    }
+
+    template <unsigned int N, typename K, typename V, class Cmp>
+    __device__ __forceinline__ void reduceKeyVal(volatile K* skeys, K& key, volatile V* svals, V& val, unsigned int tid, const Cmp& cmp)
+    {
+        reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<volatile K*, K&, volatile V*, V&, const Cmp&>(skeys, key, svals, val, tid, cmp);
+    }
+    template <unsigned int N,
+              typename K,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp>
+    __device__ __forceinline__ void reduceKeyVal(volatile K* skeys, K& key,
+                                                 const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 unsigned int tid, const Cmp& cmp)
+    {
+        reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<volatile K*, K&,
+                const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+                const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+                const Cmp&>(skeys, key, svals, val, tid, cmp);
+    }
+    template <unsigned int N,
+              typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+              typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+    __device__ __forceinline__ void reduceKeyVal(const thrust::tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                                 const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                                 const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 unsigned int tid,
+                                                 const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp)
+    {
+        reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<
+                const thrust::tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>&,
+                const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>&,
+                const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+                const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+                const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>&
+                >(skeys, key, svals, val, tid, cmp);
+    }
+
+    // smem_tuple
+
+    template <typename T0>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*>
+    smem_tuple(T0* t0)
+    {
+        return thrust::make_tuple((volatile T0*) t0);
+    }
+
+    template <typename T0, typename T1>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*>
+    smem_tuple(T0* t0, T1* t1)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1);
+    }
+
+    template <typename T0, typename T1, typename T2>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*>
+    smem_tuple(T0* t0, T1* t1, T2* t2)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*, volatile T8*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7, T8* t8)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7, (volatile T8*) t8);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8, typename T9>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*, volatile T8*, volatile T9*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7, T8* t8, T9* t9)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7, (volatile T8*) t8, (volatile T9*) t9);
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_REDUCE_HPP

3rdPart/include/opencv2/core/cuda/saturate_cast.hpp

@@ -0,0 +1,292 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_SATURATE_CAST_HPP
+#define OPENCV_CUDA_SATURATE_CAST_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(uchar v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(schar v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(ushort v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(short v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(uint v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(int v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(float v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(double v) { return _Tp(v); }
+
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(schar v)
+    {
+        uint res = 0;
+        int vi = v;
+        asm("cvt.sat.u8.s8 %0, %1;" : "=r"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(short v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.s16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(ushort v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.u16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(int v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.s32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(uint v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.u32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(float v)
+    {
+        uint res = 0;
+        asm("cvt.rni.sat.u8.f32 %0, %1;" : "=r"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        uint res = 0;
+        asm("cvt.rni.sat.u8.f64 %0, %1;" : "=r"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<uchar>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(uchar v)
+    {
+        uint res = 0;
+        uint vi = v;
+        asm("cvt.sat.s8.u8 %0, %1;" : "=r"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(short v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.s16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(ushort v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.u16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(int v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.s32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(uint v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.u32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(float v)
+    {
+        uint res = 0;
+        asm("cvt.rni.sat.s8.f32 %0, %1;" : "=r"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        uint res = 0;
+        asm("cvt.rni.sat.s8.f64 %0, %1;" : "=r"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<schar>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(schar v)
+    {
+        ushort res = 0;
+        int vi = v;
+        asm("cvt.sat.u16.s8 %0, %1;" : "=h"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(short v)
+    {
+        ushort res = 0;
+        asm("cvt.sat.u16.s16 %0, %1;" : "=h"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(int v)
+    {
+        ushort res = 0;
+        asm("cvt.sat.u16.s32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(uint v)
+    {
+        ushort res = 0;
+        asm("cvt.sat.u16.u32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(float v)
+    {
+        ushort res = 0;
+        asm("cvt.rni.sat.u16.f32 %0, %1;" : "=h"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        ushort res = 0;
+        asm("cvt.rni.sat.u16.f64 %0, %1;" : "=h"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<ushort>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ short saturate_cast<short>(ushort v)
+    {
+        short res = 0;
+        asm("cvt.sat.s16.u16 %0, %1;" : "=h"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(int v)
+    {
+        short res = 0;
+        asm("cvt.sat.s16.s32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(uint v)
+    {
+        short res = 0;
+        asm("cvt.sat.s16.u32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(float v)
+    {
+        short res = 0;
+        asm("cvt.rni.sat.s16.f32 %0, %1;" : "=h"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        short res = 0;
+        asm("cvt.rni.sat.s16.f64 %0, %1;" : "=h"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<short>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ int saturate_cast<int>(uint v)
+    {
+        int res = 0;
+        asm("cvt.sat.s32.u32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ int saturate_cast<int>(float v)
+    {
+        return __float2int_rn(v);
+    }
+    template<> __device__ __forceinline__ int saturate_cast<int>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        return __double2int_rn(v);
+    #else
+        return saturate_cast<int>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(schar v)
+    {
+        uint res = 0;
+        int vi = v;
+        asm("cvt.sat.u32.s8 %0, %1;" : "=r"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(short v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u32.s16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(int v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u32.s32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(float v)
+    {
+        return __float2uint_rn(v);
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        return __double2uint_rn(v);
+    #else
+        return saturate_cast<uint>((float)v);
+    #endif
+    }
+}}}
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_SATURATE_CAST_HPP */

3rdPart/include/opencv2/core/cuda/scan.hpp

@@ -0,0 +1,258 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_SCAN_HPP
+#define OPENCV_CUDA_SCAN_HPP
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/warp.hpp"
+#include "opencv2/core/cuda/warp_shuffle.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    enum ScanKind { EXCLUSIVE = 0,  INCLUSIVE = 1 };
+
+    template <ScanKind Kind, typename T, typename F> struct WarpScan
+    {
+        __device__ __forceinline__ WarpScan() {}
+        __device__ __forceinline__ WarpScan(const WarpScan& other) { CV_UNUSED(other); }
+
+        __device__ __forceinline__ T operator()( volatile T *ptr , const unsigned int idx)
+        {
+            const unsigned int lane = idx & 31;
+            F op;
+
+            if ( lane >=  1) ptr [idx ] = op(ptr [idx -  1], ptr [idx]);
+            if ( lane >=  2) ptr [idx ] = op(ptr [idx -  2], ptr [idx]);
+            if ( lane >=  4) ptr [idx ] = op(ptr [idx -  4], ptr [idx]);
+            if ( lane >=  8) ptr [idx ] = op(ptr [idx -  8], ptr [idx]);
+            if ( lane >= 16) ptr [idx ] = op(ptr [idx - 16], ptr [idx]);
+
+            if( Kind == INCLUSIVE )
+                return ptr [idx];
+            else
+                return (lane > 0) ? ptr [idx - 1] : 0;
+        }
+
+        __device__ __forceinline__ unsigned int index(const unsigned int tid)
+        {
+            return tid;
+        }
+
+        __device__ __forceinline__ void init(volatile T *ptr){}
+
+        static const int warp_offset      = 0;
+
+        typedef WarpScan<INCLUSIVE, T, F>  merge;
+    };
+
+    template <ScanKind Kind , typename T, typename F> struct WarpScanNoComp
+    {
+        __device__ __forceinline__ WarpScanNoComp() {}
+        __device__ __forceinline__ WarpScanNoComp(const WarpScanNoComp& other) { CV_UNUSED(other); }
+
+        __device__ __forceinline__ T operator()( volatile T *ptr , const unsigned int idx)
+        {
+            const unsigned int lane = threadIdx.x & 31;
+            F op;
+
+            ptr [idx ] = op(ptr [idx -  1], ptr [idx]);
+            ptr [idx ] = op(ptr [idx -  2], ptr [idx]);
+            ptr [idx ] = op(ptr [idx -  4], ptr [idx]);
+            ptr [idx ] = op(ptr [idx -  8], ptr [idx]);
+            ptr [idx ] = op(ptr [idx - 16], ptr [idx]);
+
+            if( Kind == INCLUSIVE )
+                return ptr [idx];
+            else
+                return (lane > 0) ? ptr [idx - 1] : 0;
+        }
+
+        __device__ __forceinline__ unsigned int index(const unsigned int tid)
+        {
+            return (tid >> warp_log) * warp_smem_stride + 16 + (tid & warp_mask);
+        }
+
+        __device__ __forceinline__ void init(volatile T *ptr)
+        {
+            ptr[threadIdx.x] = 0;
+        }
+
+        static const int warp_smem_stride = 32 + 16 + 1;
+        static const int warp_offset      = 16;
+        static const int warp_log         = 5;
+        static const int warp_mask        = 31;
+
+        typedef WarpScanNoComp<INCLUSIVE, T, F> merge;
+    };
+
+    template <ScanKind Kind , typename T, typename Sc, typename F> struct BlockScan
+    {
+        __device__ __forceinline__ BlockScan() {}
+        __device__ __forceinline__ BlockScan(const BlockScan& other) { CV_UNUSED(other); }
+
+        __device__ __forceinline__ T operator()(volatile T *ptr)
+        {
+            const unsigned int tid  = threadIdx.x;
+            const unsigned int lane = tid & warp_mask;
+            const unsigned int warp = tid >> warp_log;
+
+            Sc scan;
+            typename Sc::merge merge_scan;
+            const unsigned int idx = scan.index(tid);
+
+            T val = scan(ptr, idx);
+            __syncthreads ();
+
+            if( warp == 0)
+                scan.init(ptr);
+            __syncthreads ();
+
+            if( lane == 31 )
+                ptr [scan.warp_offset + warp ] = (Kind == INCLUSIVE) ? val : ptr [idx];
+            __syncthreads ();
+
+            if( warp == 0 )
+                merge_scan(ptr, idx);
+            __syncthreads();
+
+            if ( warp > 0)
+                val = ptr [scan.warp_offset + warp - 1] + val;
+            __syncthreads ();
+
+            ptr[idx] = val;
+            __syncthreads ();
+
+            return val ;
+        }
+
+        static const int warp_log  = 5;
+        static const int warp_mask = 31;
+    };
+
+    template <typename T>
+    __device__ T warpScanInclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+    #if __CUDA_ARCH__ >= 300
+        const unsigned int laneId = cv::cuda::device::Warp::laneId();
+
+        // scan on shuffl functions
+        #pragma unroll
+        for (int i = 1; i <= (OPENCV_CUDA_WARP_SIZE / 2); i *= 2)
+        {
+            const T n = cv::cuda::device::shfl_up(idata, i);
+            if (laneId >= i)
+                  idata += n;
+        }
+
+        return idata;
+    #else
+        unsigned int pos = 2 * tid - (tid & (OPENCV_CUDA_WARP_SIZE - 1));
+        s_Data[pos] = 0;
+        pos += OPENCV_CUDA_WARP_SIZE;
+        s_Data[pos] = idata;
+
+        s_Data[pos] += s_Data[pos - 1];
+        s_Data[pos] += s_Data[pos - 2];
+        s_Data[pos] += s_Data[pos - 4];
+        s_Data[pos] += s_Data[pos - 8];
+        s_Data[pos] += s_Data[pos - 16];
+
+        return s_Data[pos];
+    #endif
+    }
+
+    template <typename T>
+    __device__ __forceinline__ T warpScanExclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+        return warpScanInclusive(idata, s_Data, tid) - idata;
+    }
+
+    template <int tiNumScanThreads, typename T>
+    __device__ T blockScanInclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+        if (tiNumScanThreads > OPENCV_CUDA_WARP_SIZE)
+        {
+            //Bottom-level inclusive warp scan
+            T warpResult = warpScanInclusive(idata, s_Data, tid);
+
+            //Save top elements of each warp for exclusive warp scan
+            //sync to wait for warp scans to complete (because s_Data is being overwritten)
+            __syncthreads();
+            if ((tid & (OPENCV_CUDA_WARP_SIZE - 1)) == (OPENCV_CUDA_WARP_SIZE - 1))
+            {
+                s_Data[tid >> OPENCV_CUDA_LOG_WARP_SIZE] = warpResult;
+            }
+
+            //wait for warp scans to complete
+            __syncthreads();
+
+            if (tid < (tiNumScanThreads / OPENCV_CUDA_WARP_SIZE) )
+            {
+                //grab top warp elements
+                T val = s_Data[tid];
+                //calculate exclusive scan and write back to shared memory
+                s_Data[tid] = warpScanExclusive(val, s_Data, tid);
+            }
+
+            //return updated warp scans with exclusive scan results
+            __syncthreads();
+
+            return warpResult + s_Data[tid >> OPENCV_CUDA_LOG_WARP_SIZE];
+        }
+        else
+        {
+            return warpScanInclusive(idata, s_Data, tid);
+        }
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_SCAN_HPP

3rdPart/include/opencv2/core/cuda/simd_functions.hpp

@@ -0,0 +1,869 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+/*
+ * Copyright (c) 2013 NVIDIA Corporation. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ *
+ *   Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ *
+ *   Neither the name of NVIDIA Corporation nor the names of its contributors
+ *   may be used to endorse or promote products derived from this software
+ *   without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef OPENCV_CUDA_SIMD_FUNCTIONS_HPP
+#define OPENCV_CUDA_SIMD_FUNCTIONS_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    // 2
+
+    static __device__ __forceinline__ unsigned int vadd2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vadd2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vadd.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = a ^ b;          // sum bits
+        r = a + b;          // actual sum
+        s = s ^ r;          // determine carry-ins for each bit position
+        s = s & 0x00010000; // carry-in to high word (= carry-out from low word)
+        r = r - s;          // subtract out carry-out from low word
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsub2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vsub2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vsub.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = a ^ b;          // sum bits
+        r = a - b;          // actual sum
+        s = s ^ r;          // determine carry-ins for each bit position
+        s = s & 0x00010000; // borrow to high word
+        r = r + s;          // compensate for borrow from low word
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vabsdiff2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vabsdiff2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vabsdiff.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u, v;
+        s = a & 0x0000ffff; // extract low halfword
+        r = b & 0x0000ffff; // extract low halfword
+        u = ::max(r, s);    // maximum of low halfwords
+        v = ::min(r, s);    // minimum of low halfwords
+        s = a & 0xffff0000; // extract high halfword
+        r = b & 0xffff0000; // extract high halfword
+        t = ::max(r, s);    // maximum of high halfwords
+        s = ::min(r, s);    // minimum of high halfwords
+        r = u | t;          // maximum of both halfwords
+        s = v | s;          // minimum of both halfwords
+        r = r - s;          // |a - b| = max(a,b) - min(a,b);
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vavg2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+        // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+        // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+        s = a ^ b;
+        r = a & b;
+        s = s & 0xfffefffe; // ensure shift doesn't cross halfword boundaries
+        s = s >> 1;
+        s = r + s;
+
+        return s;
+    }
+
+    static __device__ __forceinline__ unsigned int vavrg2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vavrg2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+        // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+        unsigned int s;
+        s = a ^ b;
+        r = a | b;
+        s = s & 0xfffefffe; // ensure shift doesn't cross half-word boundaries
+        s = s >> 1;
+        r = r - s;
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vseteq2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r & ~c;         // msb = 1, if r was 0x0000
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpeq2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vseteq2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r & ~c;         // msb = 1, if r was 0x0000
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetge2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpge2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetge2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetgt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80008000; // msbs = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpgt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetgt2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80008000; // msbs = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetle2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmple2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetle2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetlt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmplt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetlt2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetne2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm ("vset2.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r | c;          // msb = 1, if r was not 0x0000
+        c = c & 0x80008000; // extract msbs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpne2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetne2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r | c;          // msb = 1, if r was not 0x0000
+        c = c & 0x80008000; // extract msbs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmax2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmax2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmax.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u;
+        r = a & 0x0000ffff; // extract low halfword
+        s = b & 0x0000ffff; // extract low halfword
+        t = ::max(r, s);    // maximum of low halfwords
+        r = a & 0xffff0000; // extract high halfword
+        s = b & 0xffff0000; // extract high halfword
+        u = ::max(r, s);    // maximum of high halfwords
+        r = t | u;          // combine halfword maximums
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmin2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmin2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmin.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u;
+        r = a & 0x0000ffff; // extract low halfword
+        s = b & 0x0000ffff; // extract low halfword
+        t = ::min(r, s);    // minimum of low halfwords
+        r = a & 0xffff0000; // extract high halfword
+        s = b & 0xffff0000; // extract high halfword
+        u = ::min(r, s);    // minimum of high halfwords
+        r = t | u;          // combine halfword minimums
+    #endif
+
+        return r;
+    }
+
+    // 4
+
+    static __device__ __forceinline__ unsigned int vadd4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vadd4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vadd.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t;
+        s = a ^ b;          // sum bits
+        r = a & 0x7f7f7f7f; // clear msbs
+        t = b & 0x7f7f7f7f; // clear msbs
+        s = s & 0x80808080; // msb sum bits
+        r = r + t;          // add without msbs, record carry-out in msbs
+        r = r ^ s;          // sum of msb sum and carry-in bits, w/o carry-out
+    #endif /* __CUDA_ARCH__ >= 300 */
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsub4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vsub4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vsub.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t;
+        s = a ^ ~b;         // inverted sum bits
+        r = a | 0x80808080; // set msbs
+        t = b & 0x7f7f7f7f; // clear msbs
+        s = s & 0x80808080; // inverted msb sum bits
+        r = r - t;          // subtract w/o msbs, record inverted borrows in msb
+        r = r ^ s;          // combine inverted msb sum bits and borrows
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vavg4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+        // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+        // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+        s = a ^ b;
+        r = a & b;
+        s = s & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+        s = s >> 1;
+        s = r + s;
+
+        return s;
+    }
+
+    static __device__ __forceinline__ unsigned int vavrg4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vavrg4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+        // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+        unsigned int c;
+        c = a ^ b;
+        r = a | b;
+        c = c & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+        c = c >> 1;
+        r = r - c;
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vseteq4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x80
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r & ~c;         // msb = 1, if r was 0x00
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpeq4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, t;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vseteq4(a, b);
+        t = r << 8;         // convert bool
+        r = t - r;          //  to mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        t = a ^ b;          // 0x00 if a == b
+        r = t | 0x80808080; // set msbs, to catch carry out
+        t = t ^ r;          // extract msbs, msb = 1 if t < 0x80
+        r = r - 0x01010101; // msb = 0, if t was 0x00 or 0x80
+        r = t & ~r;         // msb = 1, if t was 0x00
+        t = r >> 7;         // build mask
+        t = r - t;          //  from
+        r = t | r;          //   msbs
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetle4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmple4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetle4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80808080; // msbs = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetlt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmplt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetlt4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80808080; // msbs = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetge4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg4(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpge4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetge4(a, b);
+        s = r << 8;         // convert bool
+        r = s - r;          //  to mask
+    #else
+        asm ("not.b32 %0,%0;" : "+r"(b));
+        r = vavrg4 (a, b);  // (a + ~b + 1) / 2 = (a - b) / 2
+        r = r & 0x80808080; // msb = carry-outs
+        s = r >> 7;         // build mask
+        s = r - s;          //  from
+        r = s | r;          //   msbs
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetgt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpgt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetgt4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetne4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r | c;          // msb = 1, if r was not 0x00
+        c = c & 0x80808080; // extract msbs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpne4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetne4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r | c;          // msb = 1, if r was not 0x00
+        c = c & 0x80808080; // extract msbs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vabsdiff4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vabsdiff4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vabsdiff.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(a, b);  // mask = 0xff if a >= b
+        r = a ^ b;          //
+        s = (r &  s) ^ b;   // select a when a >= b, else select b => max(a,b)
+        r = s ^ r;          // select a when b >= a, else select b => min(a,b)
+        r = s - r;          // |a - b| = max(a,b) - min(a,b);
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmax4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmax4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmax.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(a, b);  // mask = 0xff if a >= b
+        r = a & s;          // select a when b >= a
+        s = b & ~s;         // select b when b < a
+        r = r | s;          // combine byte selections
+    #endif
+
+        return r;           // byte-wise unsigned maximum
+    }
+
+    static __device__ __forceinline__ unsigned int vmin4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmin4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmin.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(b, a);  // mask = 0xff if a >= b
+        r = a & s;          // select a when b >= a
+        s = b & ~s;         // select b when b < a
+        r = r | s;          // combine byte selections
+    #endif
+
+        return r;
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_SIMD_FUNCTIONS_HPP

3rdPart/include/opencv2/core/cuda/transform.hpp

@@ -0,0 +1,75 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TRANSFORM_HPP
+#define OPENCV_CUDA_TRANSFORM_HPP
+
+#include "common.hpp"
+#include "utility.hpp"
+#include "detail/transform_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename T, typename D, typename UnOp, typename Mask>
+    static inline void transform(PtrStepSz<T> src, PtrStepSz<D> dst, UnOp op, const Mask& mask, cudaStream_t stream)
+    {
+        typedef TransformFunctorTraits<UnOp> ft;
+        transform_detail::TransformDispatcher<VecTraits<T>::cn == 1 && VecTraits<D>::cn == 1 && ft::smart_shift != 1>::call(src, dst, op, mask, stream);
+    }
+
+    template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+    static inline void transform(PtrStepSz<T1> src1, PtrStepSz<T2> src2, PtrStepSz<D> dst, BinOp op, const Mask& mask, cudaStream_t stream)
+    {
+        typedef TransformFunctorTraits<BinOp> ft;
+        transform_detail::TransformDispatcher<VecTraits<T1>::cn == 1 && VecTraits<T2>::cn == 1 && VecTraits<D>::cn == 1 && ft::smart_shift != 1>::call(src1, src2, dst, op, mask, stream);
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TRANSFORM_HPP

3rdPart/include/opencv2/core/cuda/type_traits.hpp

@@ -0,0 +1,90 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TYPE_TRAITS_HPP
+#define OPENCV_CUDA_TYPE_TRAITS_HPP
+
+#include "detail/type_traits_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename T> struct IsSimpleParameter
+    {
+        enum {value = type_traits_detail::IsIntegral<T>::value || type_traits_detail::IsFloat<T>::value ||
+            type_traits_detail::PointerTraits<typename type_traits_detail::ReferenceTraits<T>::type>::value};
+    };
+
+    template <typename T> struct TypeTraits
+    {
+        typedef typename type_traits_detail::UnConst<T>::type                                                NonConstType;
+        typedef typename type_traits_detail::UnVolatile<T>::type                                             NonVolatileType;
+        typedef typename type_traits_detail::UnVolatile<typename type_traits_detail::UnConst<T>::type>::type UnqualifiedType;
+        typedef typename type_traits_detail::PointerTraits<UnqualifiedType>::type                            PointeeType;
+        typedef typename type_traits_detail::ReferenceTraits<T>::type                                        ReferredType;
+
+        enum { isConst          = type_traits_detail::UnConst<T>::value };
+        enum { isVolatile       = type_traits_detail::UnVolatile<T>::value };
+
+        enum { isReference      = type_traits_detail::ReferenceTraits<UnqualifiedType>::value };
+        enum { isPointer        = type_traits_detail::PointerTraits<typename type_traits_detail::ReferenceTraits<UnqualifiedType>::type>::value };
+
+        enum { isUnsignedInt    = type_traits_detail::IsUnsignedIntegral<UnqualifiedType>::value };
+        enum { isSignedInt      = type_traits_detail::IsSignedIntergral<UnqualifiedType>::value };
+        enum { isIntegral       = type_traits_detail::IsIntegral<UnqualifiedType>::value };
+        enum { isFloat          = type_traits_detail::IsFloat<UnqualifiedType>::value };
+        enum { isArith          = isIntegral || isFloat };
+        enum { isVec            = type_traits_detail::IsVec<UnqualifiedType>::value };
+
+        typedef typename type_traits_detail::Select<IsSimpleParameter<UnqualifiedType>::value,
+            T, typename type_traits_detail::AddParameterType<T>::type>::type ParameterType;
+    };
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TYPE_TRAITS_HPP

3rdPart/include/opencv2/core/cuda/utility.hpp

@@ -0,0 +1,230 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_UTILITY_HPP
+#define OPENCV_CUDA_UTILITY_HPP
+
+#include "saturate_cast.hpp"
+#include "datamov_utils.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct CV_EXPORTS ThrustAllocator
+    {
+        typedef uchar value_type;
+        virtual ~ThrustAllocator();
+        virtual __device__ __host__ uchar* allocate(size_t numBytes) = 0;
+        virtual __device__ __host__ void deallocate(uchar* ptr, size_t numBytes) = 0;
+        static ThrustAllocator& getAllocator();
+        static void setAllocator(ThrustAllocator* allocator);
+    };
+    #define OPENCV_CUDA_LOG_WARP_SIZE        (5)
+    #define OPENCV_CUDA_WARP_SIZE            (1 << OPENCV_CUDA_LOG_WARP_SIZE)
+    #define OPENCV_CUDA_LOG_MEM_BANKS        ((__CUDA_ARCH__ >= 200) ? 5 : 4) // 32 banks on fermi, 16 on tesla
+    #define OPENCV_CUDA_MEM_BANKS            (1 << OPENCV_CUDA_LOG_MEM_BANKS)
+
+    ///////////////////////////////////////////////////////////////////////////////
+    // swap
+
+    template <typename T> void __device__ __host__ __forceinline__ swap(T& a, T& b)
+    {
+        const T temp = a;
+        a = b;
+        b = temp;
+    }
+
+    ///////////////////////////////////////////////////////////////////////////////
+    // Mask Reader
+
+    struct SingleMask
+    {
+        explicit __host__ __device__ __forceinline__ SingleMask(PtrStepb mask_) : mask(mask_) {}
+        __host__ __device__ __forceinline__ SingleMask(const SingleMask& mask_): mask(mask_.mask){}
+
+        __device__ __forceinline__ bool operator()(int y, int x) const
+        {
+            return mask.ptr(y)[x] != 0;
+        }
+
+        PtrStepb mask;
+    };
+
+    struct SingleMaskChannels
+    {
+        __host__ __device__ __forceinline__ SingleMaskChannels(PtrStepb mask_, int channels_)
+        : mask(mask_), channels(channels_) {}
+        __host__ __device__ __forceinline__ SingleMaskChannels(const SingleMaskChannels& mask_)
+            :mask(mask_.mask), channels(mask_.channels){}
+
+        __device__ __forceinline__ bool operator()(int y, int x) const
+        {
+            return mask.ptr(y)[x / channels] != 0;
+        }
+
+        PtrStepb mask;
+        int channels;
+    };
+
+    struct MaskCollection
+    {
+        explicit __host__ __device__ __forceinline__ MaskCollection(PtrStepb* maskCollection_)
+            : maskCollection(maskCollection_) {}
+
+        __device__ __forceinline__ MaskCollection(const MaskCollection& masks_)
+            : maskCollection(masks_.maskCollection), curMask(masks_.curMask){}
+
+        __device__ __forceinline__ void next()
+        {
+            curMask = *maskCollection++;
+        }
+        __device__ __forceinline__ void setMask(int z)
+        {
+            curMask = maskCollection[z];
+        }
+
+        __device__ __forceinline__ bool operator()(int y, int x) const
+        {
+            uchar val;
+            return curMask.data == 0 || (ForceGlob<uchar>::Load(curMask.ptr(y), x, val), (val != 0));
+        }
+
+        const PtrStepb* maskCollection;
+        PtrStepb curMask;
+    };
+
+    struct WithOutMask
+    {
+        __host__ __device__ __forceinline__ WithOutMask(){}
+        __host__ __device__ __forceinline__ WithOutMask(const WithOutMask&){}
+
+        __device__ __forceinline__ void next() const
+        {
+        }
+        __device__ __forceinline__ void setMask(int) const
+        {
+        }
+
+        __device__ __forceinline__ bool operator()(int, int) const
+        {
+            return true;
+        }
+
+        __device__ __forceinline__ bool operator()(int, int, int) const
+        {
+            return true;
+        }
+
+        static __device__ __forceinline__ bool check(int, int)
+        {
+            return true;
+        }
+
+        static __device__ __forceinline__ bool check(int, int, int)
+        {
+            return true;
+        }
+    };
+
+    ///////////////////////////////////////////////////////////////////////////////
+    // Solve linear system
+
+    // solve 2x2 linear system Ax=b
+    template <typename T> __device__ __forceinline__ bool solve2x2(const T A[2][2], const T b[2], T x[2])
+    {
+        T det = A[0][0] * A[1][1] - A[1][0] * A[0][1];
+
+        if (det != 0)
+        {
+            double invdet = 1.0 / det;
+
+            x[0] = saturate_cast<T>(invdet * (b[0] * A[1][1] - b[1] * A[0][1]));
+
+            x[1] = saturate_cast<T>(invdet * (A[0][0] * b[1] - A[1][0] * b[0]));
+
+            return true;
+        }
+
+        return false;
+    }
+
+    // solve 3x3 linear system Ax=b
+    template <typename T> __device__ __forceinline__ bool solve3x3(const T A[3][3], const T b[3], T x[3])
+    {
+        T det = A[0][0] * (A[1][1] * A[2][2] - A[1][2] * A[2][1])
+              - A[0][1] * (A[1][0] * A[2][2] - A[1][2] * A[2][0])
+              + A[0][2] * (A[1][0] * A[2][1] - A[1][1] * A[2][0]);
+
+        if (det != 0)
+        {
+            double invdet = 1.0 / det;
+
+            x[0] = saturate_cast<T>(invdet *
+                (b[0]    * (A[1][1] * A[2][2] - A[1][2] * A[2][1]) -
+                 A[0][1] * (b[1]    * A[2][2] - A[1][2] * b[2]   ) +
+                 A[0][2] * (b[1]    * A[2][1] - A[1][1] * b[2]   )));
+
+            x[1] = saturate_cast<T>(invdet *
+                (A[0][0] * (b[1]    * A[2][2] - A[1][2] * b[2]   ) -
+                 b[0]    * (A[1][0] * A[2][2] - A[1][2] * A[2][0]) +
+                 A[0][2] * (A[1][0] * b[2]    - b[1]    * A[2][0])));
+
+            x[2] = saturate_cast<T>(invdet *
+                (A[0][0] * (A[1][1] * b[2]    - b[1]    * A[2][1]) -
+                 A[0][1] * (A[1][0] * b[2]    - b[1]    * A[2][0]) +
+                 b[0]    * (A[1][0] * A[2][1] - A[1][1] * A[2][0])));
+
+            return true;
+        }
+
+        return false;
+    }
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_UTILITY_HPP

3rdPart/include/opencv2/core/cuda/vec_distance.hpp

@@ -0,0 +1,232 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VEC_DISTANCE_HPP
+#define OPENCV_CUDA_VEC_DISTANCE_HPP
+
+#include "reduce.hpp"
+#include "functional.hpp"
+#include "detail/vec_distance_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename T> struct L1Dist
+    {
+        typedef int value_type;
+        typedef int result_type;
+
+        __device__ __forceinline__ L1Dist() : mySum(0) {}
+
+        __device__ __forceinline__ void reduceIter(int val1, int val2)
+        {
+            mySum = __sad(val1, val2, mySum);
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(int* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<int>());
+        }
+
+        __device__ __forceinline__ operator int() const
+        {
+            return mySum;
+        }
+
+        int mySum;
+    };
+    template <> struct L1Dist<float>
+    {
+        typedef float value_type;
+        typedef float result_type;
+
+        __device__ __forceinline__ L1Dist() : mySum(0.0f) {}
+
+        __device__ __forceinline__ void reduceIter(float val1, float val2)
+        {
+            mySum += ::fabs(val1 - val2);
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(float* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<float>());
+        }
+
+        __device__ __forceinline__ operator float() const
+        {
+            return mySum;
+        }
+
+        float mySum;
+    };
+
+    struct L2Dist
+    {
+        typedef float value_type;
+        typedef float result_type;
+
+        __device__ __forceinline__ L2Dist() : mySum(0.0f) {}
+
+        __device__ __forceinline__ void reduceIter(float val1, float val2)
+        {
+            float reg = val1 - val2;
+            mySum += reg * reg;
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(float* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<float>());
+        }
+
+        __device__ __forceinline__ operator float() const
+        {
+            return sqrtf(mySum);
+        }
+
+        float mySum;
+    };
+
+    struct HammingDist
+    {
+        typedef int value_type;
+        typedef int result_type;
+
+        __device__ __forceinline__ HammingDist() : mySum(0) {}
+
+        __device__ __forceinline__ void reduceIter(int val1, int val2)
+        {
+            mySum += __popc(val1 ^ val2);
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(int* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<int>());
+        }
+
+        __device__ __forceinline__ operator int() const
+        {
+            return mySum;
+        }
+
+        int mySum;
+    };
+
+    // calc distance between two vectors in global memory
+    template <int THREAD_DIM, typename Dist, typename T1, typename T2>
+    __device__ void calcVecDiffGlobal(const T1* vec1, const T2* vec2, int len, Dist& dist, typename Dist::result_type* smem, int tid)
+    {
+        for (int i = tid; i < len; i += THREAD_DIM)
+        {
+            T1 val1;
+            ForceGlob<T1>::Load(vec1, i, val1);
+
+            T2 val2;
+            ForceGlob<T2>::Load(vec2, i, val2);
+
+            dist.reduceIter(val1, val2);
+        }
+
+        dist.reduceAll<THREAD_DIM>(smem, tid);
+    }
+
+    // calc distance between two vectors, first vector is cached in register or shared memory, second vector is in global memory
+    template <int THREAD_DIM, int MAX_LEN, bool LEN_EQ_MAX_LEN, typename Dist, typename T1, typename T2>
+    __device__ __forceinline__ void calcVecDiffCached(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, typename Dist::result_type* smem, int tid)
+    {
+        vec_distance_detail::VecDiffCachedCalculator<THREAD_DIM, MAX_LEN, LEN_EQ_MAX_LEN>::calc(vecCached, vecGlob, len, dist, tid);
+
+        dist.reduceAll<THREAD_DIM>(smem, tid);
+    }
+
+    // calc distance between two vectors in global memory
+    template <int THREAD_DIM, typename T1> struct VecDiffGlobal
+    {
+        explicit __device__ __forceinline__ VecDiffGlobal(const T1* vec1_, int = 0, void* = 0, int = 0, int = 0)
+        {
+            vec1 = vec1_;
+        }
+
+        template <typename T2, typename Dist>
+        __device__ __forceinline__ void calc(const T2* vec2, int len, Dist& dist, typename Dist::result_type* smem, int tid) const
+        {
+            calcVecDiffGlobal<THREAD_DIM>(vec1, vec2, len, dist, smem, tid);
+        }
+
+        const T1* vec1;
+    };
+
+    // calc distance between two vectors, first vector is cached in register memory, second vector is in global memory
+    template <int THREAD_DIM, int MAX_LEN, bool LEN_EQ_MAX_LEN, typename U> struct VecDiffCachedRegister
+    {
+        template <typename T1> __device__ __forceinline__ VecDiffCachedRegister(const T1* vec1, int len, U* smem, int glob_tid, int tid)
+        {
+            if (glob_tid < len)
+                smem[glob_tid] = vec1[glob_tid];
+            __syncthreads();
+
+            U* vec1ValsPtr = vec1Vals;
+
+            #pragma unroll
+            for (int i = tid; i < MAX_LEN; i += THREAD_DIM)
+                *vec1ValsPtr++ = smem[i];
+
+            __syncthreads();
+        }
+
+        template <typename T2, typename Dist>
+        __device__ __forceinline__ void calc(const T2* vec2, int len, Dist& dist, typename Dist::result_type* smem, int tid) const
+        {
+            calcVecDiffCached<THREAD_DIM, MAX_LEN, LEN_EQ_MAX_LEN>(vec1Vals, vec2, len, dist, smem, tid);
+        }
+
+        U vec1Vals[MAX_LEN / THREAD_DIM];
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VEC_DISTANCE_HPP

3rdPart/include/opencv2/core/cuda/vec_math.hpp

@@ -0,0 +1,930 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VECMATH_HPP
+#define OPENCV_CUDA_VECMATH_HPP
+
+#include "vec_traits.hpp"
+#include "saturate_cast.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+
+// saturate_cast
+
+namespace vec_math_detail
+{
+    template <int cn, typename VecD> struct SatCastHelper;
+    template <typename VecD> struct SatCastHelper<1, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<2, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<3, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<4, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z), saturate_cast<D>(v.w));
+        }
+    };
+
+    template <typename VecD, typename VecS> static __device__ __forceinline__ VecD saturate_cast_helper(const VecS& v)
+    {
+        return SatCastHelper<VecTraits<VecD>::cn, VecD>::cast(v);
+    }
+}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+// unary operators
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(op (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(op (a.x), op (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(op (a.x), op (a.y), op (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(op (a.x), op (a.y), op (a.z), op (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_OP
+
+// unary functions
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x), func (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x), func (a.y), func (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x), func (a.y), func (a.z), func (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::fabsf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::fabs, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrt, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::exp, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::log, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::cosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanh, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC
+
+// binary operators (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op b.x, a.y op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op b.x, a.y op b.y, a.z op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op b.x, a.y op b.y, a.z op b.z, a.w op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_OP
+
+// binary operators (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(op, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op s); \
+    } \
+    __device__ __forceinline__ output_type ## 1 operator op(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(s op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op s, a.y op s); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(s op b.x, s op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op s, a.y op s, a.z op s); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(s op b.x, s op b.y, s op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op s, a.y op s, a.z op s, a.w op s); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(s op b.x, s op b.y, s op b.z, s op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uint, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP
+
+// binary function (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x, b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x, b.x), func (a.y, b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z), func (a.w, b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmaxf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmax, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fminf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fmin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypot, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC
+
+// binary function (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(func_name, func, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) a.x, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 1 func_name(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) s, (output_type) b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s), func ((output_type) a.w, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z), func ((output_type) s, (output_type) b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, double, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC
+
+}}} // namespace cv { namespace cuda { namespace device
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VECMATH_HPP

3rdPart/include/opencv2/core/cuda/vec_traits.hpp

@@ -0,0 +1,288 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VEC_TRAITS_HPP
+#define OPENCV_CUDA_VEC_TRAITS_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<typename T, int N> struct TypeVec;
+
+    struct __align__(8) uchar8
+    {
+        uchar a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ uchar8 make_uchar8(uchar a0, uchar a1, uchar a2, uchar a3, uchar a4, uchar a5, uchar a6, uchar a7)
+    {
+        uchar8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(8) char8
+    {
+        schar a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ char8 make_char8(schar a0, schar a1, schar a2, schar a3, schar a4, schar a5, schar a6, schar a7)
+    {
+        char8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(16) ushort8
+    {
+        ushort a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ ushort8 make_ushort8(ushort a0, ushort a1, ushort a2, ushort a3, ushort a4, ushort a5, ushort a6, ushort a7)
+    {
+        ushort8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(16) short8
+    {
+        short a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ short8 make_short8(short a0, short a1, short a2, short a3, short a4, short a5, short a6, short a7)
+    {
+        short8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(32) uint8
+    {
+        uint a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ uint8 make_uint8(uint a0, uint a1, uint a2, uint a3, uint a4, uint a5, uint a6, uint a7)
+    {
+        uint8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(32) int8
+    {
+        int a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ int8 make_int8(int a0, int a1, int a2, int a3, int a4, int a5, int a6, int a7)
+    {
+        int8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(32) float8
+    {
+        float a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ float8 make_float8(float a0, float a1, float a2, float a3, float a4, float a5, float a6, float a7)
+    {
+        float8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct double8
+    {
+        double a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ double8 make_double8(double a0, double a1, double a2, double a3, double a4, double a5, double a6, double a7)
+    {
+        double8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+
+#define OPENCV_CUDA_IMPLEMENT_TYPE_VEC(type) \
+    template<> struct TypeVec<type, 1> { typedef type vec_type; }; \
+    template<> struct TypeVec<type ## 1, 1> { typedef type ## 1 vec_type; }; \
+    template<> struct TypeVec<type, 2> { typedef type ## 2 vec_type; }; \
+    template<> struct TypeVec<type ## 2, 2> { typedef type ## 2 vec_type; }; \
+    template<> struct TypeVec<type, 3> { typedef type ## 3 vec_type; }; \
+    template<> struct TypeVec<type ## 3, 3> { typedef type ## 3 vec_type; }; \
+    template<> struct TypeVec<type, 4> { typedef type ## 4 vec_type; }; \
+    template<> struct TypeVec<type ## 4, 4> { typedef type ## 4 vec_type; }; \
+    template<> struct TypeVec<type, 8> { typedef type ## 8 vec_type; }; \
+    template<> struct TypeVec<type ## 8, 8> { typedef type ## 8 vec_type; };
+
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(uchar)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(char)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(ushort)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(short)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(int)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(uint)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(float)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(double)
+
+    #undef OPENCV_CUDA_IMPLEMENT_TYPE_VEC
+
+    template<> struct TypeVec<schar, 1> { typedef schar vec_type; };
+    template<> struct TypeVec<schar, 2> { typedef char2 vec_type; };
+    template<> struct TypeVec<schar, 3> { typedef char3 vec_type; };
+    template<> struct TypeVec<schar, 4> { typedef char4 vec_type; };
+    template<> struct TypeVec<schar, 8> { typedef char8 vec_type; };
+
+    template<> struct TypeVec<bool, 1> { typedef uchar vec_type; };
+    template<> struct TypeVec<bool, 2> { typedef uchar2 vec_type; };
+    template<> struct TypeVec<bool, 3> { typedef uchar3 vec_type; };
+    template<> struct TypeVec<bool, 4> { typedef uchar4 vec_type; };
+    template<> struct TypeVec<bool, 8> { typedef uchar8 vec_type; };
+
+    template<typename T> struct VecTraits;
+
+#define OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(type) \
+    template<> struct VecTraits<type> \
+    { \
+        typedef type elem_type; \
+        enum {cn=1}; \
+        static __device__ __host__ __forceinline__ type all(type v) {return v;} \
+        static __device__ __host__ __forceinline__ type make(type x) {return x;} \
+        static __device__ __host__ __forceinline__ type make(const type* v) {return *v;} \
+    }; \
+    template<> struct VecTraits<type ## 1> \
+    { \
+        typedef type elem_type; \
+        enum {cn=1}; \
+        static __device__ __host__ __forceinline__ type ## 1 all(type v) {return make_ ## type ## 1(v);} \
+        static __device__ __host__ __forceinline__ type ## 1 make(type x) {return make_ ## type ## 1(x);} \
+        static __device__ __host__ __forceinline__ type ## 1 make(const type* v) {return make_ ## type ## 1(*v);} \
+    }; \
+    template<> struct VecTraits<type ## 2> \
+    { \
+        typedef type elem_type; \
+        enum {cn=2}; \
+        static __device__ __host__ __forceinline__ type ## 2 all(type v) {return make_ ## type ## 2(v, v);} \
+        static __device__ __host__ __forceinline__ type ## 2 make(type x, type y) {return make_ ## type ## 2(x, y);} \
+        static __device__ __host__ __forceinline__ type ## 2 make(const type* v) {return make_ ## type ## 2(v[0], v[1]);} \
+    }; \
+    template<> struct VecTraits<type ## 3> \
+    { \
+        typedef type elem_type; \
+        enum {cn=3}; \
+        static __device__ __host__ __forceinline__ type ## 3 all(type v) {return make_ ## type ## 3(v, v, v);} \
+        static __device__ __host__ __forceinline__ type ## 3 make(type x, type y, type z) {return make_ ## type ## 3(x, y, z);} \
+        static __device__ __host__ __forceinline__ type ## 3 make(const type* v) {return make_ ## type ## 3(v[0], v[1], v[2]);} \
+    }; \
+    template<> struct VecTraits<type ## 4> \
+    { \
+        typedef type elem_type; \
+        enum {cn=4}; \
+        static __device__ __host__ __forceinline__ type ## 4 all(type v) {return make_ ## type ## 4(v, v, v, v);} \
+        static __device__ __host__ __forceinline__ type ## 4 make(type x, type y, type z, type w) {return make_ ## type ## 4(x, y, z, w);} \
+        static __device__ __host__ __forceinline__ type ## 4 make(const type* v) {return make_ ## type ## 4(v[0], v[1], v[2], v[3]);} \
+    }; \
+    template<> struct VecTraits<type ## 8> \
+    { \
+        typedef type elem_type; \
+        enum {cn=8}; \
+        static __device__ __host__ __forceinline__ type ## 8 all(type v) {return make_ ## type ## 8(v, v, v, v, v, v, v, v);} \
+        static __device__ __host__ __forceinline__ type ## 8 make(type a0, type a1, type a2, type a3, type a4, type a5, type a6, type a7) {return make_ ## type ## 8(a0, a1, a2, a3, a4, a5, a6, a7);} \
+        static __device__ __host__ __forceinline__ type ## 8 make(const type* v) {return make_ ## type ## 8(v[0], v[1], v[2], v[3], v[4], v[5], v[6], v[7]);} \
+    };
+
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(uchar)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(ushort)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(short)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(int)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(uint)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(float)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(double)
+
+    #undef OPENCV_CUDA_IMPLEMENT_VEC_TRAITS
+
+    template<> struct VecTraits<char>
+    {
+        typedef char elem_type;
+        enum {cn=1};
+        static __device__ __host__ __forceinline__ char all(char v) {return v;}
+        static __device__ __host__ __forceinline__ char make(char x) {return x;}
+        static __device__ __host__ __forceinline__ char make(const char* x) {return *x;}
+    };
+    template<> struct VecTraits<schar>
+    {
+        typedef schar elem_type;
+        enum {cn=1};
+        static __device__ __host__ __forceinline__ schar all(schar v) {return v;}
+        static __device__ __host__ __forceinline__ schar make(schar x) {return x;}
+        static __device__ __host__ __forceinline__ schar make(const schar* x) {return *x;}
+    };
+    template<> struct VecTraits<char1>
+    {
+        typedef schar elem_type;
+        enum {cn=1};
+        static __device__ __host__ __forceinline__ char1 all(schar v) {return make_char1(v);}
+        static __device__ __host__ __forceinline__ char1 make(schar x) {return make_char1(x);}
+        static __device__ __host__ __forceinline__ char1 make(const schar* v) {return make_char1(v[0]);}
+    };
+    template<> struct VecTraits<char2>
+    {
+        typedef schar elem_type;
+        enum {cn=2};
+        static __device__ __host__ __forceinline__ char2 all(schar v) {return make_char2(v, v);}
+        static __device__ __host__ __forceinline__ char2 make(schar x, schar y) {return make_char2(x, y);}
+        static __device__ __host__ __forceinline__ char2 make(const schar* v) {return make_char2(v[0], v[1]);}
+    };
+    template<> struct VecTraits<char3>
+    {
+        typedef schar elem_type;
+        enum {cn=3};
+        static __device__ __host__ __forceinline__ char3 all(schar v) {return make_char3(v, v, v);}
+        static __device__ __host__ __forceinline__ char3 make(schar x, schar y, schar z) {return make_char3(x, y, z);}
+        static __device__ __host__ __forceinline__ char3 make(const schar* v) {return make_char3(v[0], v[1], v[2]);}
+    };
+    template<> struct VecTraits<char4>
+    {
+        typedef schar elem_type;
+        enum {cn=4};
+        static __device__ __host__ __forceinline__ char4 all(schar v) {return make_char4(v, v, v, v);}
+        static __device__ __host__ __forceinline__ char4 make(schar x, schar y, schar z, schar w) {return make_char4(x, y, z, w);}
+        static __device__ __host__ __forceinline__ char4 make(const schar* v) {return make_char4(v[0], v[1], v[2], v[3]);}
+    };
+    template<> struct VecTraits<char8>
+    {
+        typedef schar elem_type;
+        enum {cn=8};
+        static __device__ __host__ __forceinline__ char8 all(schar v) {return make_char8(v, v, v, v, v, v, v, v);}
+        static __device__ __host__ __forceinline__ char8 make(schar a0, schar a1, schar a2, schar a3, schar a4, schar a5, schar a6, schar a7) {return make_char8(a0, a1, a2, a3, a4, a5, a6, a7);}
+        static __device__ __host__ __forceinline__ char8 make(const schar* v) {return make_char8(v[0], v[1], v[2], v[3], v[4], v[5], v[6], v[7]);}
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VEC_TRAITS_HPP

3rdPart/include/opencv2/core/cuda/warp.hpp

@@ -0,0 +1,139 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DEVICE_WARP_HPP
+#define OPENCV_CUDA_DEVICE_WARP_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct Warp
+    {
+        enum
+        {
+            LOG_WARP_SIZE = 5,
+            WARP_SIZE     = 1 << LOG_WARP_SIZE,
+            STRIDE        = WARP_SIZE
+        };
+
+        /** \brief Returns the warp lane ID of the calling thread. */
+        static __device__ __forceinline__ unsigned int laneId()
+        {
+            unsigned int ret;
+            asm("mov.u32 %0, %%laneid;" : "=r"(ret) );
+            return ret;
+        }
+
+        template<typename It, typename T>
+        static __device__ __forceinline__ void fill(It beg, It end, const T& value)
+        {
+            for(It t = beg + laneId(); t < end; t += STRIDE)
+                *t = value;
+        }
+
+        template<typename InIt, typename OutIt>
+        static __device__ __forceinline__ OutIt copy(InIt beg, InIt end, OutIt out)
+        {
+            for(InIt t = beg + laneId(); t < end; t += STRIDE, out += STRIDE)
+                *out = *t;
+            return out;
+        }
+
+        template<typename InIt, typename OutIt, class UnOp>
+        static __device__ __forceinline__ OutIt transform(InIt beg, InIt end, OutIt out, UnOp op)
+        {
+            for(InIt t = beg + laneId(); t < end; t += STRIDE, out += STRIDE)
+                *out = op(*t);
+            return out;
+        }
+
+        template<typename InIt1, typename InIt2, typename OutIt, class BinOp>
+        static __device__ __forceinline__ OutIt transform(InIt1 beg1, InIt1 end1, InIt2 beg2, OutIt out, BinOp op)
+        {
+            unsigned int lane = laneId();
+
+            InIt1 t1 = beg1 + lane;
+            InIt2 t2 = beg2 + lane;
+            for(; t1 < end1; t1 += STRIDE, t2 += STRIDE, out += STRIDE)
+                *out = op(*t1, *t2);
+            return out;
+        }
+
+        template <class T, class BinOp>
+        static __device__ __forceinline__ T reduce(volatile T *ptr, BinOp op)
+        {
+            const unsigned int lane = laneId();
+
+            if (lane < 16)
+            {
+                T partial = ptr[lane];
+
+                ptr[lane] = partial = op(partial, ptr[lane + 16]);
+                ptr[lane] = partial = op(partial, ptr[lane + 8]);
+                ptr[lane] = partial = op(partial, ptr[lane + 4]);
+                ptr[lane] = partial = op(partial, ptr[lane + 2]);
+                ptr[lane] = partial = op(partial, ptr[lane + 1]);
+            }
+
+            return *ptr;
+        }
+
+        template<typename OutIt, typename T>
+        static __device__ __forceinline__ void yota(OutIt beg, OutIt end, T value)
+        {
+            unsigned int lane = laneId();
+            value += lane;
+
+            for(OutIt t = beg + lane; t < end; t += STRIDE, value += STRIDE)
+                *t = value;
+        }
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_DEVICE_WARP_HPP */

3rdPart/include/opencv2/core/cuda/warp_reduce.hpp

@@ -0,0 +1,76 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_WARP_REDUCE_HPP__
+#define OPENCV_CUDA_WARP_REDUCE_HPP__
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <class T>
+    __device__ __forceinline__ T warp_reduce(volatile T *ptr , const unsigned int tid = threadIdx.x)
+    {
+        const unsigned int lane = tid & 31; // index of thread in warp (0..31)
+
+        if (lane < 16)
+        {
+            T partial = ptr[tid];
+
+            ptr[tid] = partial = partial + ptr[tid + 16];
+            ptr[tid] = partial = partial + ptr[tid + 8];
+            ptr[tid] = partial = partial + ptr[tid + 4];
+            ptr[tid] = partial = partial + ptr[tid + 2];
+            ptr[tid] = partial = partial + ptr[tid + 1];
+        }
+
+        return ptr[tid - lane];
+    }
+}}} // namespace cv { namespace cuda { namespace cudev {
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_WARP_REDUCE_HPP__ */

3rdPart/include/opencv2/core/cuda/warp_shuffle.hpp

@@ -0,0 +1,162 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_WARP_SHUFFLE_HPP
+#define OPENCV_CUDA_WARP_SHUFFLE_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+#if __CUDACC_VER_MAJOR__ >= 9
+#  define __shfl(x, y, z) __shfl_sync(0xFFFFFFFFU, x, y, z)
+#  define __shfl_up(x, y, z) __shfl_up_sync(0xFFFFFFFFU, x, y, z)
+#  define __shfl_down(x, y, z) __shfl_down_sync(0xFFFFFFFFU, x, y, z)
+#endif
+    template <typename T>
+    __device__ __forceinline__ T shfl(T val, int srcLane, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return __shfl(val, srcLane, width);
+    #else
+        return T();
+    #endif
+    }
+    __device__ __forceinline__ unsigned int shfl(unsigned int val, int srcLane, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return (unsigned int) __shfl((int) val, srcLane, width);
+    #else
+        return 0;
+    #endif
+    }
+    __device__ __forceinline__ double shfl(double val, int srcLane, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        int lo = __double2loint(val);
+        int hi = __double2hiint(val);
+
+        lo = __shfl(lo, srcLane, width);
+        hi = __shfl(hi, srcLane, width);
+
+        return __hiloint2double(hi, lo);
+    #else
+        return 0.0;
+    #endif
+    }
+
+    template <typename T>
+    __device__ __forceinline__ T shfl_down(T val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return __shfl_down(val, delta, width);
+    #else
+        return T();
+    #endif
+    }
+    __device__ __forceinline__ unsigned int shfl_down(unsigned int val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return (unsigned int) __shfl_down((int) val, delta, width);
+    #else
+        return 0;
+    #endif
+    }
+    __device__ __forceinline__ double shfl_down(double val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        int lo = __double2loint(val);
+        int hi = __double2hiint(val);
+
+        lo = __shfl_down(lo, delta, width);
+        hi = __shfl_down(hi, delta, width);
+
+        return __hiloint2double(hi, lo);
+    #else
+        return 0.0;
+    #endif
+    }
+
+    template <typename T>
+    __device__ __forceinline__ T shfl_up(T val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return __shfl_up(val, delta, width);
+    #else
+        return T();
+    #endif
+    }
+    __device__ __forceinline__ unsigned int shfl_up(unsigned int val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return (unsigned int) __shfl_up((int) val, delta, width);
+    #else
+        return 0;
+    #endif
+    }
+    __device__ __forceinline__ double shfl_up(double val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        int lo = __double2loint(val);
+        int hi = __double2hiint(val);
+
+        lo = __shfl_up(lo, delta, width);
+        hi = __shfl_up(hi, delta, width);
+
+        return __hiloint2double(hi, lo);
+    #else
+        return 0.0;
+    #endif
+    }
+}}}
+
+#  undef __shfl
+#  undef __shfl_up
+#  undef __shfl_down
+
+//! @endcond
+
+#endif // OPENCV_CUDA_WARP_SHUFFLE_HPP

3rdPart/include/opencv2/core/cuda_stream_accessor.hpp

@@ -0,0 +1,86 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CUDA_STREAM_ACCESSOR_HPP
+#define OPENCV_CORE_CUDA_STREAM_ACCESSOR_HPP
+
+#ifndef __cplusplus
+#  error cuda_stream_accessor.hpp header must be compiled as C++
+#endif
+
+/** @file cuda_stream_accessor.hpp
+ * This is only header file that depends on CUDA Runtime API. All other headers are independent.
+ */
+
+#include <cuda_runtime.h>
+#include "opencv2/core/cuda.hpp"
+
+namespace cv
+{
+    namespace cuda
+    {
+
+//! @addtogroup cudacore_struct
+//! @{
+
+        /** @brief Class that enables getting cudaStream_t from cuda::Stream
+         */
+        struct StreamAccessor
+        {
+            CV_EXPORTS static cudaStream_t getStream(const Stream& stream);
+            CV_EXPORTS static Stream wrapStream(cudaStream_t stream);
+        };
+
+        /** @brief Class that enables getting cudaEvent_t from cuda::Event
+         */
+        struct EventAccessor
+        {
+            CV_EXPORTS static cudaEvent_t getEvent(const Event& event);
+            CV_EXPORTS static Event wrapEvent(cudaEvent_t event);
+        };
+
+//! @}
+
+    }
+}
+
+#endif /* OPENCV_CORE_CUDA_STREAM_ACCESSOR_HPP */

3rdPart/include/opencv2/core/cuda_types.hpp

@@ -0,0 +1,144 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CUDA_TYPES_HPP
+#define OPENCV_CORE_CUDA_TYPES_HPP
+
+#ifndef __cplusplus
+#  error cuda_types.hpp header must be compiled as C++
+#endif
+
+#if defined(__OPENCV_BUILD) && defined(__clang__)
+#pragma clang diagnostic ignored "-Winconsistent-missing-override"
+#endif
+#if defined(__OPENCV_BUILD) && defined(__GNUC__) && __GNUC__ >= 5
+#pragma GCC diagnostic ignored "-Wsuggest-override"
+#endif
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+#ifdef __CUDACC__
+    #define __CV_CUDA_HOST_DEVICE__ __host__ __device__ __forceinline__
+#else
+    #define __CV_CUDA_HOST_DEVICE__
+#endif
+
+namespace cv
+{
+    namespace cuda
+    {
+
+        // Simple lightweight structures that encapsulates information about an image on device.
+        // It is intended to pass to nvcc-compiled code. GpuMat depends on headers that nvcc can't compile
+
+        template <typename T> struct DevPtr
+        {
+            typedef T elem_type;
+            typedef int index_type;
+
+            enum { elem_size = sizeof(elem_type) };
+
+            T* data;
+
+            __CV_CUDA_HOST_DEVICE__ DevPtr() : data(0) {}
+            __CV_CUDA_HOST_DEVICE__ DevPtr(T* data_) : data(data_) {}
+
+            __CV_CUDA_HOST_DEVICE__ size_t elemSize() const { return elem_size; }
+            __CV_CUDA_HOST_DEVICE__ operator       T*()       { return data; }
+            __CV_CUDA_HOST_DEVICE__ operator const T*() const { return data; }
+        };
+
+        template <typename T> struct PtrSz : public DevPtr<T>
+        {
+            __CV_CUDA_HOST_DEVICE__ PtrSz() : size(0) {}
+            __CV_CUDA_HOST_DEVICE__ PtrSz(T* data_, size_t size_) : DevPtr<T>(data_), size(size_) {}
+
+            size_t size;
+        };
+
+        template <typename T> struct PtrStep : public DevPtr<T>
+        {
+            __CV_CUDA_HOST_DEVICE__ PtrStep() : step(0) {}
+            __CV_CUDA_HOST_DEVICE__ PtrStep(T* data_, size_t step_) : DevPtr<T>(data_), step(step_) {}
+
+            size_t step;
+
+            __CV_CUDA_HOST_DEVICE__       T* ptr(int y = 0)       { return (      T*)( (      char*)(((DevPtr<T>*)this)->data) + y * step); }
+            __CV_CUDA_HOST_DEVICE__ const T* ptr(int y = 0) const { return (const T*)( (const char*)(((DevPtr<T>*)this)->data) + y * step); }
+
+            __CV_CUDA_HOST_DEVICE__       T& operator ()(int y, int x)       { return ptr(y)[x]; }
+            __CV_CUDA_HOST_DEVICE__ const T& operator ()(int y, int x) const { return ptr(y)[x]; }
+        };
+
+        template <typename T> struct PtrStepSz : public PtrStep<T>
+        {
+            __CV_CUDA_HOST_DEVICE__ PtrStepSz() : cols(0), rows(0) {}
+            __CV_CUDA_HOST_DEVICE__ PtrStepSz(int rows_, int cols_, T* data_, size_t step_)
+                : PtrStep<T>(data_, step_), cols(cols_), rows(rows_) {}
+
+            template <typename U>
+            explicit PtrStepSz(const PtrStepSz<U>& d) : PtrStep<T>((T*)d.data, d.step), cols(d.cols), rows(d.rows){}
+
+            int cols;
+            int rows;
+        };
+
+        typedef PtrStepSz<unsigned char> PtrStepSzb;
+        typedef PtrStepSz<unsigned short> PtrStepSzus;
+        typedef PtrStepSz<float> PtrStepSzf;
+        typedef PtrStepSz<int> PtrStepSzi;
+
+        typedef PtrStep<unsigned char> PtrStepb;
+        typedef PtrStep<unsigned short> PtrStepus;
+        typedef PtrStep<float> PtrStepf;
+        typedef PtrStep<int> PtrStepi;
+
+    }
+}
+
+//! @endcond
+
+#endif /* OPENCV_CORE_CUDA_TYPES_HPP */

3rdPart/include/opencv2/core/cv_cpu_dispatch.h

@@ -0,0 +1,350 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#if defined __OPENCV_BUILD \
+
+#include "cv_cpu_config.h"
+#include "cv_cpu_helper.h"
+
+#ifdef CV_CPU_DISPATCH_MODE
+#define CV_CPU_OPTIMIZATION_NAMESPACE __CV_CAT(opt_, CV_CPU_DISPATCH_MODE)
+#define CV_CPU_OPTIMIZATION_NAMESPACE_BEGIN namespace __CV_CAT(opt_, CV_CPU_DISPATCH_MODE) {
+#define CV_CPU_OPTIMIZATION_NAMESPACE_END }
+#else
+#define CV_CPU_OPTIMIZATION_NAMESPACE cpu_baseline
+#define CV_CPU_OPTIMIZATION_NAMESPACE_BEGIN namespace cpu_baseline {
+#define CV_CPU_OPTIMIZATION_NAMESPACE_END }
+#define CV_CPU_BASELINE_MODE 1
+#endif
+
+
+#define __CV_CPU_DISPATCH_CHAIN_END(fn, args, mode, ...)  /* done */
+#define __CV_CPU_DISPATCH(fn, args, mode, ...) __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+#define __CV_CPU_DISPATCH_EXPAND(fn, args, ...) __CV_EXPAND(__CV_CPU_DISPATCH(fn, args, __VA_ARGS__))
+#define CV_CPU_DISPATCH(fn, args, ...) __CV_CPU_DISPATCH_EXPAND(fn, args, __VA_ARGS__, END) // expand macros
+
+
+#if defined CV_ENABLE_INTRINSICS \
+    && !defined CV_DISABLE_OPTIMIZATION \
+    && !defined __CUDACC__ /* do not include SSE/AVX/NEON headers for NVCC compiler */ \
+
+#ifdef CV_CPU_COMPILE_SSE2
+#  include <emmintrin.h>
+#  define CV_MMX 1
+#  define CV_SSE 1
+#  define CV_SSE2 1
+#endif
+#ifdef CV_CPU_COMPILE_SSE3
+#  include <pmmintrin.h>
+#  define CV_SSE3 1
+#endif
+#ifdef CV_CPU_COMPILE_SSSE3
+#  include <tmmintrin.h>
+#  define CV_SSSE3 1
+#endif
+#ifdef CV_CPU_COMPILE_SSE4_1
+#  include <smmintrin.h>
+#  define CV_SSE4_1 1
+#endif
+#ifdef CV_CPU_COMPILE_SSE4_2
+#  include <nmmintrin.h>
+#  define CV_SSE4_2 1
+#endif
+#ifdef CV_CPU_COMPILE_POPCNT
+#  ifdef _MSC_VER
+#    include <nmmintrin.h>
+#    if defined(_M_X64)
+#      define CV_POPCNT_U64 _mm_popcnt_u64
+#    endif
+#    define CV_POPCNT_U32 _mm_popcnt_u32
+#  else
+#    include <popcntintrin.h>
+#    if defined(__x86_64__)
+#      define CV_POPCNT_U64 __builtin_popcountll
+#    endif
+#    define CV_POPCNT_U32 __builtin_popcount
+#  endif
+#  define CV_POPCNT 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX
+#  include <immintrin.h>
+#  define CV_AVX 1
+#endif
+#ifdef CV_CPU_COMPILE_FP16
+#  if defined(__arm__) || defined(__aarch64__) || defined(_M_ARM) || defined(_M_ARM64)
+#    include <arm_neon.h>
+#  else
+#    include <immintrin.h>
+#  endif
+#  define CV_FP16 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX2
+#  include <immintrin.h>
+#  define CV_AVX2 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX_512F
+#  include <immintrin.h>
+#  define CV_AVX_512F 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_COMMON
+#  define CV_AVX512_COMMON 1
+#  define CV_AVX_512CD 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_KNL
+#  define CV_AVX512_KNL 1
+#  define CV_AVX_512ER 1
+#  define CV_AVX_512PF 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_KNM
+#  define CV_AVX512_KNM 1
+#  define CV_AVX_5124FMAPS 1
+#  define CV_AVX_5124VNNIW 1
+#  define CV_AVX_512VPOPCNTDQ 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_SKX
+#  define CV_AVX512_SKX 1
+#  define CV_AVX_512VL 1
+#  define CV_AVX_512BW 1
+#  define CV_AVX_512DQ 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_CNL
+#  define CV_AVX512_CNL 1
+#  define CV_AVX_512IFMA 1
+#  define CV_AVX_512VBMI 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_CLX
+#  define CV_AVX512_CLX 1
+#  define CV_AVX_512VNNI 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_ICL
+#  define CV_AVX512_ICL 1
+#  undef CV_AVX_512IFMA
+#  define CV_AVX_512IFMA 1
+#  undef CV_AVX_512VBMI
+#  define CV_AVX_512VBMI 1
+#  undef CV_AVX_512VNNI
+#  define CV_AVX_512VNNI 1
+#  define CV_AVX_512VBMI2 1
+#  define CV_AVX_512BITALG 1
+#  define CV_AVX_512VPOPCNTDQ 1
+#endif
+#ifdef CV_CPU_COMPILE_FMA3
+#  define CV_FMA3 1
+#endif
+
+#if defined _WIN32 && (defined(_M_ARM) || defined(_M_ARM64)) && (defined(CV_CPU_COMPILE_NEON) || !defined(_MSC_VER))
+# include <Intrin.h>
+# include <arm_neon.h>
+# define CV_NEON 1
+#elif defined(__ARM_NEON__) || (defined (__ARM_NEON) && defined(__aarch64__))
+#  include <arm_neon.h>
+#  define CV_NEON 1
+#endif
+
+#if defined(__ARM_NEON__) || defined(__aarch64__)
+#  include <arm_neon.h>
+#endif
+
+#ifdef CV_CPU_COMPILE_VSX
+#  include <altivec.h>
+#  undef vector
+#  undef pixel
+#  undef bool
+#  define CV_VSX 1
+#endif
+
+#ifdef CV_CPU_COMPILE_VSX3
+#  define CV_VSX3 1
+#endif
+
+#ifdef CV_CPU_COMPILE_MSA
+#  include "hal/msa_macros.h"
+#  define CV_MSA 1
+#endif
+
+#ifdef __EMSCRIPTEN__
+#  define CV_WASM_SIMD 1
+#  include <wasm_simd128.h>
+#endif
+
+#endif // CV_ENABLE_INTRINSICS && !CV_DISABLE_OPTIMIZATION && !__CUDACC__
+
+#if defined CV_CPU_COMPILE_AVX && !defined CV_CPU_BASELINE_COMPILE_AVX
+struct VZeroUpperGuard {
+#ifdef __GNUC__
+    __attribute__((always_inline))
+#endif
+    inline VZeroUpperGuard() { _mm256_zeroupper(); }
+#ifdef __GNUC__
+    __attribute__((always_inline))
+#endif
+    inline ~VZeroUpperGuard() { _mm256_zeroupper(); }
+};
+#define __CV_AVX_GUARD VZeroUpperGuard __vzeroupper_guard; CV_UNUSED(__vzeroupper_guard);
+#endif
+
+#ifdef __CV_AVX_GUARD
+#define CV_AVX_GUARD __CV_AVX_GUARD
+#else
+#define CV_AVX_GUARD
+#endif
+
+#endif // __OPENCV_BUILD
+
+
+
+#if !defined __OPENCV_BUILD /* Compatibility code */ \
+    && !defined __CUDACC__ /* do not include SSE/AVX/NEON headers for NVCC compiler */
+#if defined __SSE2__ || defined _M_X64 || (defined _M_IX86_FP && _M_IX86_FP >= 2)
+#  include <emmintrin.h>
+#  define CV_MMX 1
+#  define CV_SSE 1
+#  define CV_SSE2 1
+#elif defined _WIN32 && (defined(_M_ARM) || defined(_M_ARM64)) && (defined(CV_CPU_COMPILE_NEON) || !defined(_MSC_VER))
+# include <Intrin.h>
+# include <arm_neon.h>
+# define CV_NEON 1
+#elif defined(__ARM_NEON__) || (defined (__ARM_NEON) && defined(__aarch64__))
+#  include <arm_neon.h>
+#  define CV_NEON 1
+#elif defined(__VSX__) && defined(__PPC64__) && defined(__LITTLE_ENDIAN__)
+#  include <altivec.h>
+#  undef vector
+#  undef pixel
+#  undef bool
+#  define CV_VSX 1
+#endif
+
+#ifdef __F16C__
+#  include <immintrin.h>
+#  define CV_FP16 1
+#endif
+
+#endif // !__OPENCV_BUILD && !__CUDACC (Compatibility code)
+
+
+
+#ifndef CV_MMX
+#  define CV_MMX 0
+#endif
+#ifndef CV_SSE
+#  define CV_SSE 0
+#endif
+#ifndef CV_SSE2
+#  define CV_SSE2 0
+#endif
+#ifndef CV_SSE3
+#  define CV_SSE3 0
+#endif
+#ifndef CV_SSSE3
+#  define CV_SSSE3 0
+#endif
+#ifndef CV_SSE4_1
+#  define CV_SSE4_1 0
+#endif
+#ifndef CV_SSE4_2
+#  define CV_SSE4_2 0
+#endif
+#ifndef CV_POPCNT
+#  define CV_POPCNT 0
+#endif
+#ifndef CV_AVX
+#  define CV_AVX 0
+#endif
+#ifndef CV_FP16
+#  define CV_FP16 0
+#endif
+#ifndef CV_AVX2
+#  define CV_AVX2 0
+#endif
+#ifndef CV_FMA3
+#  define CV_FMA3 0
+#endif
+#ifndef CV_AVX_512F
+#  define CV_AVX_512F 0
+#endif
+#ifndef CV_AVX_512BW
+#  define CV_AVX_512BW 0
+#endif
+#ifndef CV_AVX_512CD
+#  define CV_AVX_512CD 0
+#endif
+#ifndef CV_AVX_512DQ
+#  define CV_AVX_512DQ 0
+#endif
+#ifndef CV_AVX_512ER
+#  define CV_AVX_512ER 0
+#endif
+#ifndef CV_AVX_512IFMA
+#  define CV_AVX_512IFMA 0
+#endif
+#define CV_AVX_512IFMA512 CV_AVX_512IFMA // deprecated
+#ifndef CV_AVX_512PF
+#  define CV_AVX_512PF 0
+#endif
+#ifndef CV_AVX_512VBMI
+#  define CV_AVX_512VBMI 0
+#endif
+#ifndef CV_AVX_512VL
+#  define CV_AVX_512VL 0
+#endif
+#ifndef CV_AVX_5124FMAPS
+#  define CV_AVX_5124FMAPS 0
+#endif
+#ifndef CV_AVX_5124VNNIW
+#  define CV_AVX_5124VNNIW 0
+#endif
+#ifndef CV_AVX_512VPOPCNTDQ
+#  define CV_AVX_512VPOPCNTDQ 0
+#endif
+#ifndef CV_AVX_512VNNI
+#  define CV_AVX_512VNNI 0
+#endif
+#ifndef CV_AVX_512VBMI2
+#  define CV_AVX_512VBMI2 0
+#endif
+#ifndef CV_AVX_512BITALG
+#  define CV_AVX_512BITALG 0
+#endif
+#ifndef CV_AVX512_COMMON
+#  define CV_AVX512_COMMON 0
+#endif
+#ifndef CV_AVX512_KNL
+#  define CV_AVX512_KNL 0
+#endif
+#ifndef CV_AVX512_KNM
+#  define CV_AVX512_KNM 0
+#endif
+#ifndef CV_AVX512_SKX
+#  define CV_AVX512_SKX 0
+#endif
+#ifndef CV_AVX512_CNL
+#  define CV_AVX512_CNL 0
+#endif
+#ifndef CV_AVX512_CLX
+#  define CV_AVX512_CLX 0
+#endif
+#ifndef CV_AVX512_ICL
+#  define CV_AVX512_ICL 0
+#endif
+
+#ifndef CV_NEON
+#  define CV_NEON 0
+#endif
+
+#ifndef CV_VSX
+#  define CV_VSX 0
+#endif
+
+#ifndef CV_VSX3
+#  define CV_VSX3 0
+#endif
+
+#ifndef CV_MSA
+#  define CV_MSA 0
+#endif
+
+#ifndef CV_WASM_SIMD
+#  define CV_WASM_SIMD 0
+#endif

3rdPart/include/opencv2/core/cv_cpu_helper.h

@@ -0,0 +1,487 @@
+// AUTOGENERATED, DO NOT EDIT
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE
+#  define CV_TRY_SSE 1
+#  define CV_CPU_FORCE_SSE 1
+#  define CV_CPU_HAS_SUPPORT_SSE 1
+#  define CV_CPU_CALL_SSE(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE_(fn, args) return (opt_SSE::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE
+#  define CV_TRY_SSE 1
+#  define CV_CPU_FORCE_SSE 0
+#  define CV_CPU_HAS_SUPPORT_SSE (cv::checkHardwareSupport(CV_CPU_SSE))
+#  define CV_CPU_CALL_SSE(fn, args) if (CV_CPU_HAS_SUPPORT_SSE) return (opt_SSE::fn args)
+#  define CV_CPU_CALL_SSE_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE) return (opt_SSE::fn args)
+#else
+#  define CV_TRY_SSE 0
+#  define CV_CPU_FORCE_SSE 0
+#  define CV_CPU_HAS_SUPPORT_SSE 0
+#  define CV_CPU_CALL_SSE(fn, args)
+#  define CV_CPU_CALL_SSE_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE(fn, args, mode, ...)  CV_CPU_CALL_SSE(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE2
+#  define CV_TRY_SSE2 1
+#  define CV_CPU_FORCE_SSE2 1
+#  define CV_CPU_HAS_SUPPORT_SSE2 1
+#  define CV_CPU_CALL_SSE2(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE2_(fn, args) return (opt_SSE2::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE2
+#  define CV_TRY_SSE2 1
+#  define CV_CPU_FORCE_SSE2 0
+#  define CV_CPU_HAS_SUPPORT_SSE2 (cv::checkHardwareSupport(CV_CPU_SSE2))
+#  define CV_CPU_CALL_SSE2(fn, args) if (CV_CPU_HAS_SUPPORT_SSE2) return (opt_SSE2::fn args)
+#  define CV_CPU_CALL_SSE2_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE2) return (opt_SSE2::fn args)
+#else
+#  define CV_TRY_SSE2 0
+#  define CV_CPU_FORCE_SSE2 0
+#  define CV_CPU_HAS_SUPPORT_SSE2 0
+#  define CV_CPU_CALL_SSE2(fn, args)
+#  define CV_CPU_CALL_SSE2_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE2(fn, args, mode, ...)  CV_CPU_CALL_SSE2(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE3
+#  define CV_TRY_SSE3 1
+#  define CV_CPU_FORCE_SSE3 1
+#  define CV_CPU_HAS_SUPPORT_SSE3 1
+#  define CV_CPU_CALL_SSE3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE3_(fn, args) return (opt_SSE3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE3
+#  define CV_TRY_SSE3 1
+#  define CV_CPU_FORCE_SSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSE3 (cv::checkHardwareSupport(CV_CPU_SSE3))
+#  define CV_CPU_CALL_SSE3(fn, args) if (CV_CPU_HAS_SUPPORT_SSE3) return (opt_SSE3::fn args)
+#  define CV_CPU_CALL_SSE3_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE3) return (opt_SSE3::fn args)
+#else
+#  define CV_TRY_SSE3 0
+#  define CV_CPU_FORCE_SSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSE3 0
+#  define CV_CPU_CALL_SSE3(fn, args)
+#  define CV_CPU_CALL_SSE3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE3(fn, args, mode, ...)  CV_CPU_CALL_SSE3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSSE3
+#  define CV_TRY_SSSE3 1
+#  define CV_CPU_FORCE_SSSE3 1
+#  define CV_CPU_HAS_SUPPORT_SSSE3 1
+#  define CV_CPU_CALL_SSSE3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSSE3_(fn, args) return (opt_SSSE3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSSE3
+#  define CV_TRY_SSSE3 1
+#  define CV_CPU_FORCE_SSSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSSE3 (cv::checkHardwareSupport(CV_CPU_SSSE3))
+#  define CV_CPU_CALL_SSSE3(fn, args) if (CV_CPU_HAS_SUPPORT_SSSE3) return (opt_SSSE3::fn args)
+#  define CV_CPU_CALL_SSSE3_(fn, args) if (CV_CPU_HAS_SUPPORT_SSSE3) return (opt_SSSE3::fn args)
+#else
+#  define CV_TRY_SSSE3 0
+#  define CV_CPU_FORCE_SSSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSSE3 0
+#  define CV_CPU_CALL_SSSE3(fn, args)
+#  define CV_CPU_CALL_SSSE3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSSE3(fn, args, mode, ...)  CV_CPU_CALL_SSSE3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE4_1
+#  define CV_TRY_SSE4_1 1
+#  define CV_CPU_FORCE_SSE4_1 1
+#  define CV_CPU_HAS_SUPPORT_SSE4_1 1
+#  define CV_CPU_CALL_SSE4_1(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE4_1_(fn, args) return (opt_SSE4_1::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE4_1
+#  define CV_TRY_SSE4_1 1
+#  define CV_CPU_FORCE_SSE4_1 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_1 (cv::checkHardwareSupport(CV_CPU_SSE4_1))
+#  define CV_CPU_CALL_SSE4_1(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_1) return (opt_SSE4_1::fn args)
+#  define CV_CPU_CALL_SSE4_1_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_1) return (opt_SSE4_1::fn args)
+#else
+#  define CV_TRY_SSE4_1 0
+#  define CV_CPU_FORCE_SSE4_1 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_1 0
+#  define CV_CPU_CALL_SSE4_1(fn, args)
+#  define CV_CPU_CALL_SSE4_1_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE4_1(fn, args, mode, ...)  CV_CPU_CALL_SSE4_1(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE4_2
+#  define CV_TRY_SSE4_2 1
+#  define CV_CPU_FORCE_SSE4_2 1
+#  define CV_CPU_HAS_SUPPORT_SSE4_2 1
+#  define CV_CPU_CALL_SSE4_2(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE4_2_(fn, args) return (opt_SSE4_2::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE4_2
+#  define CV_TRY_SSE4_2 1
+#  define CV_CPU_FORCE_SSE4_2 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_2 (cv::checkHardwareSupport(CV_CPU_SSE4_2))
+#  define CV_CPU_CALL_SSE4_2(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_2) return (opt_SSE4_2::fn args)
+#  define CV_CPU_CALL_SSE4_2_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_2) return (opt_SSE4_2::fn args)
+#else
+#  define CV_TRY_SSE4_2 0
+#  define CV_CPU_FORCE_SSE4_2 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_2 0
+#  define CV_CPU_CALL_SSE4_2(fn, args)
+#  define CV_CPU_CALL_SSE4_2_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE4_2(fn, args, mode, ...)  CV_CPU_CALL_SSE4_2(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_POPCNT
+#  define CV_TRY_POPCNT 1
+#  define CV_CPU_FORCE_POPCNT 1
+#  define CV_CPU_HAS_SUPPORT_POPCNT 1
+#  define CV_CPU_CALL_POPCNT(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_POPCNT_(fn, args) return (opt_POPCNT::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_POPCNT
+#  define CV_TRY_POPCNT 1
+#  define CV_CPU_FORCE_POPCNT 0
+#  define CV_CPU_HAS_SUPPORT_POPCNT (cv::checkHardwareSupport(CV_CPU_POPCNT))
+#  define CV_CPU_CALL_POPCNT(fn, args) if (CV_CPU_HAS_SUPPORT_POPCNT) return (opt_POPCNT::fn args)
+#  define CV_CPU_CALL_POPCNT_(fn, args) if (CV_CPU_HAS_SUPPORT_POPCNT) return (opt_POPCNT::fn args)
+#else
+#  define CV_TRY_POPCNT 0
+#  define CV_CPU_FORCE_POPCNT 0
+#  define CV_CPU_HAS_SUPPORT_POPCNT 0
+#  define CV_CPU_CALL_POPCNT(fn, args)
+#  define CV_CPU_CALL_POPCNT_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_POPCNT(fn, args, mode, ...)  CV_CPU_CALL_POPCNT(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX
+#  define CV_TRY_AVX 1
+#  define CV_CPU_FORCE_AVX 1
+#  define CV_CPU_HAS_SUPPORT_AVX 1
+#  define CV_CPU_CALL_AVX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX_(fn, args) return (opt_AVX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX
+#  define CV_TRY_AVX 1
+#  define CV_CPU_FORCE_AVX 0
+#  define CV_CPU_HAS_SUPPORT_AVX (cv::checkHardwareSupport(CV_CPU_AVX))
+#  define CV_CPU_CALL_AVX(fn, args) if (CV_CPU_HAS_SUPPORT_AVX) return (opt_AVX::fn args)
+#  define CV_CPU_CALL_AVX_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX) return (opt_AVX::fn args)
+#else
+#  define CV_TRY_AVX 0
+#  define CV_CPU_FORCE_AVX 0
+#  define CV_CPU_HAS_SUPPORT_AVX 0
+#  define CV_CPU_CALL_AVX(fn, args)
+#  define CV_CPU_CALL_AVX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX(fn, args, mode, ...)  CV_CPU_CALL_AVX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_FP16
+#  define CV_TRY_FP16 1
+#  define CV_CPU_FORCE_FP16 1
+#  define CV_CPU_HAS_SUPPORT_FP16 1
+#  define CV_CPU_CALL_FP16(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_FP16_(fn, args) return (opt_FP16::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_FP16
+#  define CV_TRY_FP16 1
+#  define CV_CPU_FORCE_FP16 0
+#  define CV_CPU_HAS_SUPPORT_FP16 (cv::checkHardwareSupport(CV_CPU_FP16))
+#  define CV_CPU_CALL_FP16(fn, args) if (CV_CPU_HAS_SUPPORT_FP16) return (opt_FP16::fn args)
+#  define CV_CPU_CALL_FP16_(fn, args) if (CV_CPU_HAS_SUPPORT_FP16) return (opt_FP16::fn args)
+#else
+#  define CV_TRY_FP16 0
+#  define CV_CPU_FORCE_FP16 0
+#  define CV_CPU_HAS_SUPPORT_FP16 0
+#  define CV_CPU_CALL_FP16(fn, args)
+#  define CV_CPU_CALL_FP16_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_FP16(fn, args, mode, ...)  CV_CPU_CALL_FP16(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX2
+#  define CV_TRY_AVX2 1
+#  define CV_CPU_FORCE_AVX2 1
+#  define CV_CPU_HAS_SUPPORT_AVX2 1
+#  define CV_CPU_CALL_AVX2(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX2_(fn, args) return (opt_AVX2::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX2
+#  define CV_TRY_AVX2 1
+#  define CV_CPU_FORCE_AVX2 0
+#  define CV_CPU_HAS_SUPPORT_AVX2 (cv::checkHardwareSupport(CV_CPU_AVX2))
+#  define CV_CPU_CALL_AVX2(fn, args) if (CV_CPU_HAS_SUPPORT_AVX2) return (opt_AVX2::fn args)
+#  define CV_CPU_CALL_AVX2_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX2) return (opt_AVX2::fn args)
+#else
+#  define CV_TRY_AVX2 0
+#  define CV_CPU_FORCE_AVX2 0
+#  define CV_CPU_HAS_SUPPORT_AVX2 0
+#  define CV_CPU_CALL_AVX2(fn, args)
+#  define CV_CPU_CALL_AVX2_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX2(fn, args, mode, ...)  CV_CPU_CALL_AVX2(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_FMA3
+#  define CV_TRY_FMA3 1
+#  define CV_CPU_FORCE_FMA3 1
+#  define CV_CPU_HAS_SUPPORT_FMA3 1
+#  define CV_CPU_CALL_FMA3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_FMA3_(fn, args) return (opt_FMA3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_FMA3
+#  define CV_TRY_FMA3 1
+#  define CV_CPU_FORCE_FMA3 0
+#  define CV_CPU_HAS_SUPPORT_FMA3 (cv::checkHardwareSupport(CV_CPU_FMA3))
+#  define CV_CPU_CALL_FMA3(fn, args) if (CV_CPU_HAS_SUPPORT_FMA3) return (opt_FMA3::fn args)
+#  define CV_CPU_CALL_FMA3_(fn, args) if (CV_CPU_HAS_SUPPORT_FMA3) return (opt_FMA3::fn args)
+#else
+#  define CV_TRY_FMA3 0
+#  define CV_CPU_FORCE_FMA3 0
+#  define CV_CPU_HAS_SUPPORT_FMA3 0
+#  define CV_CPU_CALL_FMA3(fn, args)
+#  define CV_CPU_CALL_FMA3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_FMA3(fn, args, mode, ...)  CV_CPU_CALL_FMA3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX_512F
+#  define CV_TRY_AVX_512F 1
+#  define CV_CPU_FORCE_AVX_512F 1
+#  define CV_CPU_HAS_SUPPORT_AVX_512F 1
+#  define CV_CPU_CALL_AVX_512F(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX_512F_(fn, args) return (opt_AVX_512F::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX_512F
+#  define CV_TRY_AVX_512F 1
+#  define CV_CPU_FORCE_AVX_512F 0
+#  define CV_CPU_HAS_SUPPORT_AVX_512F (cv::checkHardwareSupport(CV_CPU_AVX_512F))
+#  define CV_CPU_CALL_AVX_512F(fn, args) if (CV_CPU_HAS_SUPPORT_AVX_512F) return (opt_AVX_512F::fn args)
+#  define CV_CPU_CALL_AVX_512F_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX_512F) return (opt_AVX_512F::fn args)
+#else
+#  define CV_TRY_AVX_512F 0
+#  define CV_CPU_FORCE_AVX_512F 0
+#  define CV_CPU_HAS_SUPPORT_AVX_512F 0
+#  define CV_CPU_CALL_AVX_512F(fn, args)
+#  define CV_CPU_CALL_AVX_512F_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX_512F(fn, args, mode, ...)  CV_CPU_CALL_AVX_512F(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_COMMON
+#  define CV_TRY_AVX512_COMMON 1
+#  define CV_CPU_FORCE_AVX512_COMMON 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_COMMON 1
+#  define CV_CPU_CALL_AVX512_COMMON(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_COMMON_(fn, args) return (opt_AVX512_COMMON::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_COMMON
+#  define CV_TRY_AVX512_COMMON 1
+#  define CV_CPU_FORCE_AVX512_COMMON 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_COMMON (cv::checkHardwareSupport(CV_CPU_AVX512_COMMON))
+#  define CV_CPU_CALL_AVX512_COMMON(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_COMMON) return (opt_AVX512_COMMON::fn args)
+#  define CV_CPU_CALL_AVX512_COMMON_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_COMMON) return (opt_AVX512_COMMON::fn args)
+#else
+#  define CV_TRY_AVX512_COMMON 0
+#  define CV_CPU_FORCE_AVX512_COMMON 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_COMMON 0
+#  define CV_CPU_CALL_AVX512_COMMON(fn, args)
+#  define CV_CPU_CALL_AVX512_COMMON_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_COMMON(fn, args, mode, ...)  CV_CPU_CALL_AVX512_COMMON(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_KNL
+#  define CV_TRY_AVX512_KNL 1
+#  define CV_CPU_FORCE_AVX512_KNL 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNL 1
+#  define CV_CPU_CALL_AVX512_KNL(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_KNL_(fn, args) return (opt_AVX512_KNL::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_KNL
+#  define CV_TRY_AVX512_KNL 1
+#  define CV_CPU_FORCE_AVX512_KNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNL (cv::checkHardwareSupport(CV_CPU_AVX512_KNL))
+#  define CV_CPU_CALL_AVX512_KNL(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNL) return (opt_AVX512_KNL::fn args)
+#  define CV_CPU_CALL_AVX512_KNL_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNL) return (opt_AVX512_KNL::fn args)
+#else
+#  define CV_TRY_AVX512_KNL 0
+#  define CV_CPU_FORCE_AVX512_KNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNL 0
+#  define CV_CPU_CALL_AVX512_KNL(fn, args)
+#  define CV_CPU_CALL_AVX512_KNL_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_KNL(fn, args, mode, ...)  CV_CPU_CALL_AVX512_KNL(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_KNM
+#  define CV_TRY_AVX512_KNM 1
+#  define CV_CPU_FORCE_AVX512_KNM 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNM 1
+#  define CV_CPU_CALL_AVX512_KNM(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_KNM_(fn, args) return (opt_AVX512_KNM::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_KNM
+#  define CV_TRY_AVX512_KNM 1
+#  define CV_CPU_FORCE_AVX512_KNM 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNM (cv::checkHardwareSupport(CV_CPU_AVX512_KNM))
+#  define CV_CPU_CALL_AVX512_KNM(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNM) return (opt_AVX512_KNM::fn args)
+#  define CV_CPU_CALL_AVX512_KNM_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNM) return (opt_AVX512_KNM::fn args)
+#else
+#  define CV_TRY_AVX512_KNM 0
+#  define CV_CPU_FORCE_AVX512_KNM 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNM 0
+#  define CV_CPU_CALL_AVX512_KNM(fn, args)
+#  define CV_CPU_CALL_AVX512_KNM_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_KNM(fn, args, mode, ...)  CV_CPU_CALL_AVX512_KNM(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_SKX
+#  define CV_TRY_AVX512_SKX 1
+#  define CV_CPU_FORCE_AVX512_SKX 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_SKX 1
+#  define CV_CPU_CALL_AVX512_SKX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_SKX_(fn, args) return (opt_AVX512_SKX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_SKX
+#  define CV_TRY_AVX512_SKX 1
+#  define CV_CPU_FORCE_AVX512_SKX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_SKX (cv::checkHardwareSupport(CV_CPU_AVX512_SKX))
+#  define CV_CPU_CALL_AVX512_SKX(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_SKX) return (opt_AVX512_SKX::fn args)
+#  define CV_CPU_CALL_AVX512_SKX_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_SKX) return (opt_AVX512_SKX::fn args)
+#else
+#  define CV_TRY_AVX512_SKX 0
+#  define CV_CPU_FORCE_AVX512_SKX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_SKX 0
+#  define CV_CPU_CALL_AVX512_SKX(fn, args)
+#  define CV_CPU_CALL_AVX512_SKX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_SKX(fn, args, mode, ...)  CV_CPU_CALL_AVX512_SKX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_CNL
+#  define CV_TRY_AVX512_CNL 1
+#  define CV_CPU_FORCE_AVX512_CNL 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_CNL 1
+#  define CV_CPU_CALL_AVX512_CNL(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_CNL_(fn, args) return (opt_AVX512_CNL::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_CNL
+#  define CV_TRY_AVX512_CNL 1
+#  define CV_CPU_FORCE_AVX512_CNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CNL (cv::checkHardwareSupport(CV_CPU_AVX512_CNL))
+#  define CV_CPU_CALL_AVX512_CNL(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CNL) return (opt_AVX512_CNL::fn args)
+#  define CV_CPU_CALL_AVX512_CNL_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CNL) return (opt_AVX512_CNL::fn args)
+#else
+#  define CV_TRY_AVX512_CNL 0
+#  define CV_CPU_FORCE_AVX512_CNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CNL 0
+#  define CV_CPU_CALL_AVX512_CNL(fn, args)
+#  define CV_CPU_CALL_AVX512_CNL_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_CNL(fn, args, mode, ...)  CV_CPU_CALL_AVX512_CNL(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_CLX
+#  define CV_TRY_AVX512_CLX 1
+#  define CV_CPU_FORCE_AVX512_CLX 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_CLX 1
+#  define CV_CPU_CALL_AVX512_CLX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_CLX_(fn, args) return (opt_AVX512_CLX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_CLX
+#  define CV_TRY_AVX512_CLX 1
+#  define CV_CPU_FORCE_AVX512_CLX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CLX (cv::checkHardwareSupport(CV_CPU_AVX512_CLX))
+#  define CV_CPU_CALL_AVX512_CLX(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CLX) return (opt_AVX512_CLX::fn args)
+#  define CV_CPU_CALL_AVX512_CLX_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CLX) return (opt_AVX512_CLX::fn args)
+#else
+#  define CV_TRY_AVX512_CLX 0
+#  define CV_CPU_FORCE_AVX512_CLX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CLX 0
+#  define CV_CPU_CALL_AVX512_CLX(fn, args)
+#  define CV_CPU_CALL_AVX512_CLX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_CLX(fn, args, mode, ...)  CV_CPU_CALL_AVX512_CLX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_ICL
+#  define CV_TRY_AVX512_ICL 1
+#  define CV_CPU_FORCE_AVX512_ICL 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_ICL 1
+#  define CV_CPU_CALL_AVX512_ICL(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_ICL_(fn, args) return (opt_AVX512_ICL::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_ICL
+#  define CV_TRY_AVX512_ICL 1
+#  define CV_CPU_FORCE_AVX512_ICL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_ICL (cv::checkHardwareSupport(CV_CPU_AVX512_ICL))
+#  define CV_CPU_CALL_AVX512_ICL(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_ICL) return (opt_AVX512_ICL::fn args)
+#  define CV_CPU_CALL_AVX512_ICL_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_ICL) return (opt_AVX512_ICL::fn args)
+#else
+#  define CV_TRY_AVX512_ICL 0
+#  define CV_CPU_FORCE_AVX512_ICL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_ICL 0
+#  define CV_CPU_CALL_AVX512_ICL(fn, args)
+#  define CV_CPU_CALL_AVX512_ICL_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_ICL(fn, args, mode, ...)  CV_CPU_CALL_AVX512_ICL(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_NEON
+#  define CV_TRY_NEON 1
+#  define CV_CPU_FORCE_NEON 1
+#  define CV_CPU_HAS_SUPPORT_NEON 1
+#  define CV_CPU_CALL_NEON(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_NEON_(fn, args) return (opt_NEON::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_NEON
+#  define CV_TRY_NEON 1
+#  define CV_CPU_FORCE_NEON 0
+#  define CV_CPU_HAS_SUPPORT_NEON (cv::checkHardwareSupport(CV_CPU_NEON))
+#  define CV_CPU_CALL_NEON(fn, args) if (CV_CPU_HAS_SUPPORT_NEON) return (opt_NEON::fn args)
+#  define CV_CPU_CALL_NEON_(fn, args) if (CV_CPU_HAS_SUPPORT_NEON) return (opt_NEON::fn args)
+#else
+#  define CV_TRY_NEON 0
+#  define CV_CPU_FORCE_NEON 0
+#  define CV_CPU_HAS_SUPPORT_NEON 0
+#  define CV_CPU_CALL_NEON(fn, args)
+#  define CV_CPU_CALL_NEON_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_NEON(fn, args, mode, ...)  CV_CPU_CALL_NEON(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_MSA
+#  define CV_TRY_MSA 1
+#  define CV_CPU_FORCE_MSA 1
+#  define CV_CPU_HAS_SUPPORT_MSA 1
+#  define CV_CPU_CALL_MSA(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_MSA_(fn, args) return (opt_MSA::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_MSA
+#  define CV_TRY_MSA 1
+#  define CV_CPU_FORCE_MSA 0
+#  define CV_CPU_HAS_SUPPORT_MSA (cv::checkHardwareSupport(CV_CPU_MSA))
+#  define CV_CPU_CALL_MSA(fn, args) if (CV_CPU_HAS_SUPPORT_MSA) return (opt_MSA::fn args)
+#  define CV_CPU_CALL_MSA_(fn, args) if (CV_CPU_HAS_SUPPORT_MSA) return (opt_MSA::fn args)
+#else
+#  define CV_TRY_MSA 0
+#  define CV_CPU_FORCE_MSA 0
+#  define CV_CPU_HAS_SUPPORT_MSA 0
+#  define CV_CPU_CALL_MSA(fn, args)
+#  define CV_CPU_CALL_MSA_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_MSA(fn, args, mode, ...)  CV_CPU_CALL_MSA(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_VSX
+#  define CV_TRY_VSX 1
+#  define CV_CPU_FORCE_VSX 1
+#  define CV_CPU_HAS_SUPPORT_VSX 1
+#  define CV_CPU_CALL_VSX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_VSX_(fn, args) return (opt_VSX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_VSX
+#  define CV_TRY_VSX 1
+#  define CV_CPU_FORCE_VSX 0
+#  define CV_CPU_HAS_SUPPORT_VSX (cv::checkHardwareSupport(CV_CPU_VSX))
+#  define CV_CPU_CALL_VSX(fn, args) if (CV_CPU_HAS_SUPPORT_VSX) return (opt_VSX::fn args)
+#  define CV_CPU_CALL_VSX_(fn, args) if (CV_CPU_HAS_SUPPORT_VSX) return (opt_VSX::fn args)
+#else
+#  define CV_TRY_VSX 0
+#  define CV_CPU_FORCE_VSX 0
+#  define CV_CPU_HAS_SUPPORT_VSX 0
+#  define CV_CPU_CALL_VSX(fn, args)
+#  define CV_CPU_CALL_VSX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_VSX(fn, args, mode, ...)  CV_CPU_CALL_VSX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_VSX3
+#  define CV_TRY_VSX3 1
+#  define CV_CPU_FORCE_VSX3 1
+#  define CV_CPU_HAS_SUPPORT_VSX3 1
+#  define CV_CPU_CALL_VSX3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_VSX3_(fn, args) return (opt_VSX3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_VSX3
+#  define CV_TRY_VSX3 1
+#  define CV_CPU_FORCE_VSX3 0
+#  define CV_CPU_HAS_SUPPORT_VSX3 (cv::checkHardwareSupport(CV_CPU_VSX3))
+#  define CV_CPU_CALL_VSX3(fn, args) if (CV_CPU_HAS_SUPPORT_VSX3) return (opt_VSX3::fn args)
+#  define CV_CPU_CALL_VSX3_(fn, args) if (CV_CPU_HAS_SUPPORT_VSX3) return (opt_VSX3::fn args)
+#else
+#  define CV_TRY_VSX3 0
+#  define CV_CPU_FORCE_VSX3 0
+#  define CV_CPU_HAS_SUPPORT_VSX3 0
+#  define CV_CPU_CALL_VSX3(fn, args)
+#  define CV_CPU_CALL_VSX3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_VSX3(fn, args, mode, ...)  CV_CPU_CALL_VSX3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#define CV_CPU_CALL_BASELINE(fn, args) return (cpu_baseline::fn args)
+#define __CV_CPU_DISPATCH_CHAIN_BASELINE(fn, args, mode, ...)  CV_CPU_CALL_BASELINE(fn, args) /* last in sequence */

3rdPart/include/opencv2/core/cvdef.h

@@ -0,0 +1,849 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CVDEF_H
+#define OPENCV_CORE_CVDEF_H
+
+#include "opencv2/core/version.hpp"
+
+//! @addtogroup core_utils
+//! @{
+
+#ifdef OPENCV_INCLUDE_PORT_FILE  // User-provided header file with custom platform configuration
+#include OPENCV_INCLUDE_PORT_FILE
+#endif
+
+#if !defined CV_DOXYGEN && !defined CV_IGNORE_DEBUG_BUILD_GUARD
+#if (defined(_MSC_VER) && (defined(DEBUG) || defined(_DEBUG))) || \
+    (defined(_GLIBCXX_DEBUG) || defined(_GLIBCXX_DEBUG_PEDANTIC))
+// Guard to prevent using of binary incompatible binaries / runtimes
+// https://github.com/opencv/opencv/pull/9161
+#define CV__DEBUG_NS_BEGIN namespace debug_build_guard {
+#define CV__DEBUG_NS_END }
+namespace cv { namespace debug_build_guard { } using namespace debug_build_guard; }
+#endif
+#endif
+
+#ifndef CV__DEBUG_NS_BEGIN
+#define CV__DEBUG_NS_BEGIN
+#define CV__DEBUG_NS_END
+#endif
+
+
+#ifdef __OPENCV_BUILD
+#include "cvconfig.h"
+#endif
+
+#ifndef __CV_EXPAND
+#define __CV_EXPAND(x) x
+#endif
+
+#ifndef __CV_CAT
+#define __CV_CAT__(x, y) x ## y
+#define __CV_CAT_(x, y) __CV_CAT__(x, y)
+#define __CV_CAT(x, y) __CV_CAT_(x, y)
+#endif
+
+#define __CV_VA_NUM_ARGS_HELPER(_1, _2, _3, _4, _5, _6, _7, _8, _9, _10, N, ...) N
+#define __CV_VA_NUM_ARGS(...) __CV_VA_NUM_ARGS_HELPER(__VA_ARGS__, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0)
+
+#ifdef CV_Func
+// keep current value (through OpenCV port file)
+#elif defined __GNUC__ || (defined (__cpluscplus) && (__cpluscplus >= 201103))
+#define CV_Func __func__
+#elif defined __clang__ && (__clang_minor__ * 100 + __clang_major__ >= 305)
+#define CV_Func __func__
+#elif defined(__STDC_VERSION__) && (__STDC_VERSION >= 199901)
+#define CV_Func __func__
+#elif defined _MSC_VER
+#define CV_Func __FUNCTION__
+#elif defined(__INTEL_COMPILER) && (_INTEL_COMPILER >= 600)
+#define CV_Func __FUNCTION__
+#elif defined __IBMCPP__ && __IBMCPP__ >=500
+#define CV_Func __FUNCTION__
+#elif defined __BORLAND__ && (__BORLANDC__ >= 0x550)
+#define CV_Func __FUNC__
+#else
+#define CV_Func "<unknown>"
+#endif
+
+//! @cond IGNORED
+
+//////////////// static assert /////////////////
+#define CVAUX_CONCAT_EXP(a, b) a##b
+#define CVAUX_CONCAT(a, b) CVAUX_CONCAT_EXP(a,b)
+
+#if defined(__clang__)
+#  ifndef __has_extension
+#    define __has_extension __has_feature /* compatibility, for older versions of clang */
+#  endif
+#  if __has_extension(cxx_static_assert)
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  elif __has_extension(c_static_assert)
+#    define CV_StaticAssert(condition, reason)    _Static_assert((condition), reason " " #condition)
+#  endif
+#elif defined(__GNUC__)
+#  if (defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L)
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  endif
+#elif defined(_MSC_VER)
+#  if _MSC_VER >= 1600 /* MSVC 10 */
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  endif
+#endif
+#ifndef CV_StaticAssert
+#  if !defined(__clang__) && defined(__GNUC__) && (__GNUC__*100 + __GNUC_MINOR__ > 302)
+#    define CV_StaticAssert(condition, reason) ({ extern int __attribute__((error("CV_StaticAssert: " reason " " #condition))) CV_StaticAssert(); ((condition) ? 0 : CV_StaticAssert()); })
+#  else
+namespace cv {
+     template <bool x> struct CV_StaticAssert_failed;
+     template <> struct CV_StaticAssert_failed<true> { enum { val = 1 }; };
+     template<int x> struct CV_StaticAssert_test {};
+}
+#    define CV_StaticAssert(condition, reason)\
+       typedef cv::CV_StaticAssert_test< sizeof(cv::CV_StaticAssert_failed< static_cast<bool>(condition) >) > CVAUX_CONCAT(CV_StaticAssert_failed_at_, __LINE__)
+#  endif
+#endif
+
+// Suppress warning "-Wdeprecated-declarations" / C4996
+#if defined(_MSC_VER)
+    #define CV_DO_PRAGMA(x) __pragma(x)
+#elif defined(__GNUC__)
+    #define CV_DO_PRAGMA(x) _Pragma (#x)
+#else
+    #define CV_DO_PRAGMA(x)
+#endif
+
+#ifdef _MSC_VER
+#define CV_SUPPRESS_DEPRECATED_START \
+    CV_DO_PRAGMA(warning(push)) \
+    CV_DO_PRAGMA(warning(disable: 4996))
+#define CV_SUPPRESS_DEPRECATED_END CV_DO_PRAGMA(warning(pop))
+#elif defined (__clang__) || ((__GNUC__)  && (__GNUC__*100 + __GNUC_MINOR__ > 405))
+#define CV_SUPPRESS_DEPRECATED_START \
+    CV_DO_PRAGMA(GCC diagnostic push) \
+    CV_DO_PRAGMA(GCC diagnostic ignored "-Wdeprecated-declarations")
+#define CV_SUPPRESS_DEPRECATED_END CV_DO_PRAGMA(GCC diagnostic pop)
+#else
+#define CV_SUPPRESS_DEPRECATED_START
+#define CV_SUPPRESS_DEPRECATED_END
+#endif
+
+#define CV_UNUSED(name) (void)name
+
+#if defined __GNUC__ && !defined __EXCEPTIONS
+#define CV_TRY
+#define CV_CATCH(A, B) for (A B; false; )
+#define CV_CATCH_ALL if (false)
+#define CV_THROW(A) abort()
+#define CV_RETHROW() abort()
+#else
+#define CV_TRY try
+#define CV_CATCH(A, B) catch(const A & B)
+#define CV_CATCH_ALL catch(...)
+#define CV_THROW(A) throw A
+#define CV_RETHROW() throw
+#endif
+
+//! @endcond
+
+// undef problematic defines sometimes defined by system headers (windows.h in particular)
+#undef small
+#undef min
+#undef max
+#undef abs
+#undef Complex
+
+#if defined __cplusplus
+#include <limits>
+#else
+#include <limits.h>
+#endif
+
+#include "opencv2/core/hal/interface.h"
+
+#if defined __ICL
+#  define CV_ICC   __ICL
+#elif defined __ICC
+#  define CV_ICC   __ICC
+#elif defined __ECL
+#  define CV_ICC   __ECL
+#elif defined __ECC
+#  define CV_ICC   __ECC
+#elif defined __INTEL_COMPILER
+#  define CV_ICC   __INTEL_COMPILER
+#endif
+
+#ifndef CV_INLINE
+#  if defined __cplusplus
+#    define CV_INLINE static inline
+#  elif defined _MSC_VER
+#    define CV_INLINE __inline
+#  else
+#    define CV_INLINE static
+#  endif
+#endif
+
+#ifndef CV_ALWAYS_INLINE
+#if defined(__GNUC__) && (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1))
+#define CV_ALWAYS_INLINE inline __attribute__((always_inline))
+#elif defined(_MSC_VER)
+#define CV_ALWAYS_INLINE __forceinline
+#else
+#define CV_ALWAYS_INLINE inline
+#endif
+#endif
+
+#if defined CV_DISABLE_OPTIMIZATION || (defined CV_ICC && !defined CV_ENABLE_UNROLLED)
+#  define CV_ENABLE_UNROLLED 0
+#else
+#  define CV_ENABLE_UNROLLED 1
+#endif
+
+#ifdef __GNUC__
+#  define CV_DECL_ALIGNED(x) __attribute__ ((aligned (x)))
+#elif defined _MSC_VER
+#  define CV_DECL_ALIGNED(x) __declspec(align(x))
+#else
+#  define CV_DECL_ALIGNED(x)
+#endif
+
+/* CPU features and intrinsics support */
+#define CV_CPU_NONE             0
+#define CV_CPU_MMX              1
+#define CV_CPU_SSE              2
+#define CV_CPU_SSE2             3
+#define CV_CPU_SSE3             4
+#define CV_CPU_SSSE3            5
+#define CV_CPU_SSE4_1           6
+#define CV_CPU_SSE4_2           7
+#define CV_CPU_POPCNT           8
+#define CV_CPU_FP16             9
+#define CV_CPU_AVX              10
+#define CV_CPU_AVX2             11
+#define CV_CPU_FMA3             12
+
+#define CV_CPU_AVX_512F         13
+#define CV_CPU_AVX_512BW        14
+#define CV_CPU_AVX_512CD        15
+#define CV_CPU_AVX_512DQ        16
+#define CV_CPU_AVX_512ER        17
+#define CV_CPU_AVX_512IFMA512   18 // deprecated
+#define CV_CPU_AVX_512IFMA      18
+#define CV_CPU_AVX_512PF        19
+#define CV_CPU_AVX_512VBMI      20
+#define CV_CPU_AVX_512VL        21
+#define CV_CPU_AVX_512VBMI2     22
+#define CV_CPU_AVX_512VNNI      23
+#define CV_CPU_AVX_512BITALG    24
+#define CV_CPU_AVX_512VPOPCNTDQ 25
+#define CV_CPU_AVX_5124VNNIW    26
+#define CV_CPU_AVX_5124FMAPS    27
+
+#define CV_CPU_NEON             100
+
+#define CV_CPU_MSA              150
+
+#define CV_CPU_VSX              200
+#define CV_CPU_VSX3             201
+
+// CPU features groups
+#define CV_CPU_AVX512_SKX       256
+#define CV_CPU_AVX512_COMMON    257
+#define CV_CPU_AVX512_KNL       258
+#define CV_CPU_AVX512_KNM       259
+#define CV_CPU_AVX512_CNL       260
+#define CV_CPU_AVX512_CLX       261
+#define CV_CPU_AVX512_ICL       262
+
+// when adding to this list remember to update the following enum
+#define CV_HARDWARE_MAX_FEATURE 512
+
+/** @brief Available CPU features.
+*/
+enum CpuFeatures {
+    CPU_MMX             = 1,
+    CPU_SSE             = 2,
+    CPU_SSE2            = 3,
+    CPU_SSE3            = 4,
+    CPU_SSSE3           = 5,
+    CPU_SSE4_1          = 6,
+    CPU_SSE4_2          = 7,
+    CPU_POPCNT          = 8,
+    CPU_FP16            = 9,
+    CPU_AVX             = 10,
+    CPU_AVX2            = 11,
+    CPU_FMA3            = 12,
+
+    CPU_AVX_512F        = 13,
+    CPU_AVX_512BW       = 14,
+    CPU_AVX_512CD       = 15,
+    CPU_AVX_512DQ       = 16,
+    CPU_AVX_512ER       = 17,
+    CPU_AVX_512IFMA512  = 18, // deprecated
+    CPU_AVX_512IFMA     = 18,
+    CPU_AVX_512PF       = 19,
+    CPU_AVX_512VBMI     = 20,
+    CPU_AVX_512VL       = 21,
+    CPU_AVX_512VBMI2    = 22,
+    CPU_AVX_512VNNI     = 23,
+    CPU_AVX_512BITALG   = 24,
+    CPU_AVX_512VPOPCNTDQ= 25,
+    CPU_AVX_5124VNNIW   = 26,
+    CPU_AVX_5124FMAPS   = 27,
+
+    CPU_NEON            = 100,
+
+    CPU_MSA             = 150,
+
+    CPU_VSX             = 200,
+    CPU_VSX3            = 201,
+
+    CPU_AVX512_SKX      = 256, //!< Skylake-X with AVX-512F/CD/BW/DQ/VL
+    CPU_AVX512_COMMON   = 257, //!< Common instructions AVX-512F/CD for all CPUs that support AVX-512
+    CPU_AVX512_KNL      = 258, //!< Knights Landing with AVX-512F/CD/ER/PF
+    CPU_AVX512_KNM      = 259, //!< Knights Mill with AVX-512F/CD/ER/PF/4FMAPS/4VNNIW/VPOPCNTDQ
+    CPU_AVX512_CNL      = 260, //!< Cannon Lake with AVX-512F/CD/BW/DQ/VL/IFMA/VBMI
+    CPU_AVX512_CLX      = 261, //!< Cascade Lake with AVX-512F/CD/BW/DQ/VL/VNNI
+    CPU_AVX512_ICL      = 262, //!< Ice Lake with AVX-512F/CD/BW/DQ/VL/IFMA/VBMI/VNNI/VBMI2/BITALG/VPOPCNTDQ
+
+    CPU_MAX_FEATURE     = 512  // see CV_HARDWARE_MAX_FEATURE
+};
+
+
+#include "cv_cpu_dispatch.h"
+
+#if !defined(CV_STRONG_ALIGNMENT) && defined(__arm__) && !(defined(__aarch64__) || defined(_M_ARM64))
+// int*, int64* should be propertly aligned pointers on ARMv7
+#define CV_STRONG_ALIGNMENT 1
+#endif
+#if !defined(CV_STRONG_ALIGNMENT)
+#define CV_STRONG_ALIGNMENT 0
+#endif
+
+/* fundamental constants */
+#define CV_PI   3.1415926535897932384626433832795
+#define CV_2PI  6.283185307179586476925286766559
+#define CV_LOG2 0.69314718055994530941723212145818
+
+#if defined __ARM_FP16_FORMAT_IEEE \
+    && !defined __CUDACC__
+#  define CV_FP16_TYPE 1
+#else
+#  define CV_FP16_TYPE 0
+#endif
+
+typedef union Cv16suf
+{
+    short i;
+    ushort u;
+#if CV_FP16_TYPE
+    __fp16 h;
+#endif
+}
+Cv16suf;
+
+typedef union Cv32suf
+{
+    int i;
+    unsigned u;
+    float f;
+}
+Cv32suf;
+
+typedef union Cv64suf
+{
+    int64 i;
+    uint64 u;
+    double f;
+}
+Cv64suf;
+
+#ifndef OPENCV_ABI_COMPATIBILITY
+#define OPENCV_ABI_COMPATIBILITY 300
+#endif
+
+#ifdef __OPENCV_BUILD
+#  define DISABLE_OPENCV_24_COMPATIBILITY
+#  define OPENCV_DISABLE_DEPRECATED_COMPATIBILITY
+#endif
+
+#ifndef CV_EXPORTS
+# if (defined _WIN32 || defined WINCE || defined __CYGWIN__) && defined(CVAPI_EXPORTS)
+#   define CV_EXPORTS __declspec(dllexport)
+# elif defined __GNUC__ && __GNUC__ >= 4 && (defined(CVAPI_EXPORTS) || defined(__APPLE__))
+#   define CV_EXPORTS __attribute__ ((visibility ("default")))
+# endif
+#endif
+
+#ifndef CV_EXPORTS
+# define CV_EXPORTS
+#endif
+
+#ifdef _MSC_VER
+#   define CV_EXPORTS_TEMPLATE
+#else
+#   define CV_EXPORTS_TEMPLATE CV_EXPORTS
+#endif
+
+#ifndef CV_DEPRECATED
+#  if defined(__GNUC__)
+#    define CV_DEPRECATED __attribute__ ((deprecated))
+#  elif defined(_MSC_VER)
+#    define CV_DEPRECATED __declspec(deprecated)
+#  else
+#    define CV_DEPRECATED
+#  endif
+#endif
+
+#ifndef CV_DEPRECATED_EXTERNAL
+#  if defined(__OPENCV_BUILD)
+#    define CV_DEPRECATED_EXTERNAL /* nothing */
+#  else
+#    define CV_DEPRECATED_EXTERNAL CV_DEPRECATED
+#  endif
+#endif
+
+
+#ifndef CV_EXTERN_C
+#  ifdef __cplusplus
+#    define CV_EXTERN_C extern "C"
+#  else
+#    define CV_EXTERN_C
+#  endif
+#endif
+
+/* special informative macros for wrapper generators */
+#define CV_EXPORTS_W CV_EXPORTS
+#define CV_EXPORTS_W_SIMPLE CV_EXPORTS
+#define CV_EXPORTS_AS(synonym) CV_EXPORTS
+#define CV_EXPORTS_W_MAP CV_EXPORTS
+#define CV_IN_OUT
+#define CV_OUT
+#define CV_PROP
+#define CV_PROP_RW
+#define CV_WRAP
+#define CV_WRAP_AS(synonym)
+
+/****************************************************************************************\
+*                                  Matrix type (Mat)                                     *
+\****************************************************************************************/
+
+#define CV_MAT_CN_MASK          ((CV_CN_MAX - 1) << CV_CN_SHIFT)
+#define CV_MAT_CN(flags)        ((((flags) & CV_MAT_CN_MASK) >> CV_CN_SHIFT) + 1)
+#define CV_MAT_TYPE_MASK        (CV_DEPTH_MAX*CV_CN_MAX - 1)
+#define CV_MAT_TYPE(flags)      ((flags) & CV_MAT_TYPE_MASK)
+#define CV_MAT_CONT_FLAG_SHIFT  14
+#define CV_MAT_CONT_FLAG        (1 << CV_MAT_CONT_FLAG_SHIFT)
+#define CV_IS_MAT_CONT(flags)   ((flags) & CV_MAT_CONT_FLAG)
+#define CV_IS_CONT_MAT          CV_IS_MAT_CONT
+#define CV_SUBMAT_FLAG_SHIFT    15
+#define CV_SUBMAT_FLAG          (1 << CV_SUBMAT_FLAG_SHIFT)
+#define CV_IS_SUBMAT(flags)     ((flags) & CV_MAT_SUBMAT_FLAG)
+
+/** Size of each channel item,
+   0x8442211 = 1000 0100 0100 0010 0010 0001 0001 ~ array of sizeof(arr_type_elem) */
+#define CV_ELEM_SIZE1(type) \
+    ((((sizeof(size_t)<<28)|0x8442211) >> CV_MAT_DEPTH(type)*4) & 15)
+
+/** 0x3a50 = 11 10 10 01 01 00 00 ~ array of log2(sizeof(arr_type_elem)) */
+#define CV_ELEM_SIZE(type) \
+    (CV_MAT_CN(type) << ((((sizeof(size_t)/4+1)*16384|0x3a50) >> CV_MAT_DEPTH(type)*2) & 3))
+
+#ifndef MIN
+#  define MIN(a,b)  ((a) > (b) ? (b) : (a))
+#endif
+
+#ifndef MAX
+#  define MAX(a,b)  ((a) < (b) ? (b) : (a))
+#endif
+
+/****************************************************************************************\
+*                                    static analysys                                     *
+\****************************************************************************************/
+
+// In practice, some macro are not processed correctly (noreturn is not detected).
+// We need to use simplified definition for them.
+#ifndef CV_STATIC_ANALYSIS
+# if defined(__KLOCWORK__) || defined(__clang_analyzer__) || defined(__COVERITY__)
+#   define CV_STATIC_ANALYSIS 1
+# endif
+#else
+# if defined(CV_STATIC_ANALYSIS) && !(__CV_CAT(1, CV_STATIC_ANALYSIS) == 1)  // defined and not empty
+#   if 0 == CV_STATIC_ANALYSIS
+#     undef CV_STATIC_ANALYSIS
+#   endif
+# endif
+#endif
+
+/****************************************************************************************\
+*                                    Thread sanitizer                                    *
+\****************************************************************************************/
+#ifndef CV_THREAD_SANITIZER
+# if defined(__has_feature)
+#   if __has_feature(thread_sanitizer)
+#     define CV_THREAD_SANITIZER
+#   endif
+# endif
+#endif
+
+/****************************************************************************************\
+*          exchange-add operation for atomic operations on reference counters            *
+\****************************************************************************************/
+
+#ifdef CV_XADD
+  // allow to use user-defined macro
+#elif defined __GNUC__ || defined __clang__
+#  if defined __clang__ && __clang_major__ >= 3 && !defined __ANDROID__ && !defined __EMSCRIPTEN__ && !defined(__CUDACC__)  && !defined __INTEL_COMPILER
+#    ifdef __ATOMIC_ACQ_REL
+#      define CV_XADD(addr, delta) __c11_atomic_fetch_add((_Atomic(int)*)(addr), delta, __ATOMIC_ACQ_REL)
+#    else
+#      define CV_XADD(addr, delta) __atomic_fetch_add((_Atomic(int)*)(addr), delta, 4)
+#    endif
+#  else
+#    if defined __ATOMIC_ACQ_REL && !defined __clang__
+       // version for gcc >= 4.7
+#      define CV_XADD(addr, delta) (int)__atomic_fetch_add((unsigned*)(addr), (unsigned)(delta), __ATOMIC_ACQ_REL)
+#    else
+#      define CV_XADD(addr, delta) (int)__sync_fetch_and_add((unsigned*)(addr), (unsigned)(delta))
+#    endif
+#  endif
+#elif defined _MSC_VER && !defined RC_INVOKED
+#  include <intrin.h>
+#  define CV_XADD(addr, delta) (int)_InterlockedExchangeAdd((long volatile*)addr, delta)
+#else
+  #ifdef OPENCV_FORCE_UNSAFE_XADD
+    CV_INLINE int CV_XADD(int* addr, int delta) { int tmp = *addr; *addr += delta; return tmp; }
+  #else
+    #error "OpenCV: can't define safe CV_XADD macro for current platform (unsupported). Define CV_XADD macro through custom port header (see OPENCV_INCLUDE_PORT_FILE)"
+  #endif
+#endif
+
+
+/****************************************************************************************\
+*                                  CV_NORETURN attribute                                 *
+\****************************************************************************************/
+
+#ifndef CV_NORETURN
+#  if defined(__GNUC__)
+#    define CV_NORETURN __attribute__((__noreturn__))
+#  elif defined(_MSC_VER) && (_MSC_VER >= 1300)
+#    define CV_NORETURN __declspec(noreturn)
+#  else
+#    define CV_NORETURN /* nothing by default */
+#  endif
+#endif
+
+
+/****************************************************************************************\
+*                                  CV_NODISCARD attribute                                *
+* encourages the compiler to issue a warning if the return value is discarded (C++17)    *
+\****************************************************************************************/
+#ifndef CV_NODISCARD
+#  if defined(__GNUC__)
+#    define CV_NODISCARD __attribute__((__warn_unused_result__)) // at least available with GCC 3.4
+#  elif defined(__clang__) && defined(__has_attribute)
+#    if __has_attribute(__warn_unused_result__)
+#      define CV_NODISCARD __attribute__((__warn_unused_result__))
+#    endif
+#  endif
+#endif
+#ifndef CV_NODISCARD
+#  define CV_NODISCARD /* nothing by default */
+#endif
+
+
+/****************************************************************************************\
+*                                    C++ 11                                              *
+\****************************************************************************************/
+#ifndef CV_CXX11
+#  if __cplusplus >= 201103L || (defined(_MSC_VER) && _MSC_VER >= 1800)
+#    define CV_CXX11 1
+#  endif
+#else
+#  if CV_CXX11 == 0
+#    undef CV_CXX11
+#  endif
+#endif
+
+
+/****************************************************************************************\
+*                                    C++ Move semantics                                  *
+\****************************************************************************************/
+
+#ifndef CV_CXX_MOVE_SEMANTICS
+#  if __cplusplus >= 201103L || defined(__GXX_EXPERIMENTAL_CXX0X__) || (defined(_MSC_VER) && _MSC_VER >= 1600)
+#    define CV_CXX_MOVE_SEMANTICS 1
+#  elif defined(__clang)
+#    if __has_feature(cxx_rvalue_references)
+#      define CV_CXX_MOVE_SEMANTICS 1
+#    endif
+#  endif
+#else
+#  if CV_CXX_MOVE_SEMANTICS == 0
+#    undef CV_CXX_MOVE_SEMANTICS
+#  endif
+#endif
+
+#ifdef CV_CXX_MOVE_SEMANTICS
+#define CV_CXX_MOVE(x) std::move(x)
+#else
+#define CV_CXX_MOVE(x) (x)
+#endif
+
+
+/****************************************************************************************\
+*                                    C++11 std::array                                    *
+\****************************************************************************************/
+
+#ifndef CV_CXX_STD_ARRAY
+#  if __cplusplus >= 201103L || (defined(__cplusplus) && defined(_MSC_VER) && _MSC_VER >= 1900/*MSVS 2015*/)
+#    define CV_CXX_STD_ARRAY 1
+#    include <array>
+#  endif
+#else
+#  if CV_CXX_STD_ARRAY == 0
+#    undef CV_CXX_STD_ARRAY
+#  endif
+#endif
+
+
+/****************************************************************************************\
+*                                 C++11 override / final                                 *
+\****************************************************************************************/
+
+#ifndef CV_OVERRIDE
+#  ifdef CV_CXX11
+#    define CV_OVERRIDE override
+#  endif
+#endif
+#ifndef CV_OVERRIDE
+#  define CV_OVERRIDE
+#endif
+
+#ifndef CV_FINAL
+#  ifdef CV_CXX11
+#    define CV_FINAL final
+#  endif
+#endif
+#ifndef CV_FINAL
+#  define CV_FINAL
+#endif
+
+/****************************************************************************************\
+*                                     C++11 noexcept                                     *
+\****************************************************************************************/
+
+#ifndef CV_NOEXCEPT
+#  if __cplusplus >= 201103L || (defined(_MSC_VER) && _MSC_VER >= 1900/*MSVS 2015*/)
+#    define CV_NOEXCEPT noexcept
+#  endif
+#endif
+#ifndef CV_NOEXCEPT
+#  define CV_NOEXCEPT
+#endif
+
+
+
+// Integer types portatibility
+#ifdef OPENCV_STDINT_HEADER
+#include OPENCV_STDINT_HEADER
+#elif defined(__cplusplus)
+#if defined(_MSC_VER) && _MSC_VER < 1600 /* MSVS 2010 */
+namespace cv {
+typedef signed char int8_t;
+typedef unsigned char uint8_t;
+typedef signed short int16_t;
+typedef unsigned short uint16_t;
+typedef signed int int32_t;
+typedef unsigned int uint32_t;
+typedef signed __int64 int64_t;
+typedef unsigned __int64 uint64_t;
+}
+#elif defined(_MSC_VER) || __cplusplus >= 201103L
+#include <cstdint>
+namespace cv {
+using std::int8_t;
+using std::uint8_t;
+using std::int16_t;
+using std::uint16_t;
+using std::int32_t;
+using std::uint32_t;
+using std::int64_t;
+using std::uint64_t;
+}
+#else
+#include <stdint.h>
+namespace cv {
+typedef ::int8_t int8_t;
+typedef ::uint8_t uint8_t;
+typedef ::int16_t int16_t;
+typedef ::uint16_t uint16_t;
+typedef ::int32_t int32_t;
+typedef ::uint32_t uint32_t;
+typedef ::int64_t int64_t;
+typedef ::uint64_t uint64_t;
+}
+#endif
+#else // pure C
+#include <stdint.h>
+#endif
+
+#ifdef __cplusplus
+namespace cv
+{
+
+class float16_t
+{
+public:
+#if CV_FP16_TYPE
+
+    float16_t() {}
+    explicit float16_t(float x) { h = (__fp16)x; }
+    operator float() const { return (float)h; }
+    static float16_t fromBits(ushort w)
+    {
+        Cv16suf u;
+        u.u = w;
+        float16_t result;
+        result.h = u.h;
+        return result;
+    }
+    static float16_t zero()
+    {
+        float16_t result;
+        result.h = (__fp16)0;
+        return result;
+    }
+    ushort bits() const
+    {
+        Cv16suf u;
+        u.h = h;
+        return u.u;
+    }
+protected:
+    __fp16 h;
+
+#else
+    float16_t() {}
+    explicit float16_t(float x)
+    {
+    #if CV_FP16
+        __m128 v = _mm_load_ss(&x);
+        w = (ushort)_mm_cvtsi128_si32(_mm_cvtps_ph(v, 0));
+    #else
+        Cv32suf in;
+        in.f = x;
+        unsigned sign = in.u & 0x80000000;
+        in.u ^= sign;
+
+        if( in.u >= 0x47800000 )
+            w = (ushort)(in.u > 0x7f800000 ? 0x7e00 : 0x7c00);
+        else
+        {
+            if (in.u < 0x38800000)
+            {
+                in.f += 0.5f;
+                w = (ushort)(in.u - 0x3f000000);
+            }
+            else
+            {
+                unsigned t = in.u + 0xc8000fff;
+                w = (ushort)((t + ((in.u >> 13) & 1)) >> 13);
+            }
+        }
+
+        w = (ushort)(w | (sign >> 16));
+    #endif
+    }
+
+    operator float() const
+    {
+    #if CV_FP16
+        float f;
+        _mm_store_ss(&f, _mm_cvtph_ps(_mm_cvtsi32_si128(w)));
+        return f;
+    #else
+        Cv32suf out;
+
+        unsigned t = ((w & 0x7fff) << 13) + 0x38000000;
+        unsigned sign = (w & 0x8000) << 16;
+        unsigned e = w & 0x7c00;
+
+        out.u = t + (1 << 23);
+        out.u = (e >= 0x7c00 ? t + 0x38000000 :
+                 e == 0 ? (out.f -= 6.103515625e-05f, out.u) : t) | sign;
+        return out.f;
+    #endif
+    }
+
+    static float16_t fromBits(ushort b)
+    {
+        float16_t result;
+        result.w = b;
+        return result;
+    }
+    static float16_t zero()
+    {
+        float16_t result;
+        result.w = (ushort)0;
+        return result;
+    }
+    ushort bits() const { return w; }
+protected:
+    ushort w;
+
+#endif
+};
+
+}
+#endif
+
+//! @}
+
+#ifndef __cplusplus
+#include "opencv2/core/fast_math.hpp" // define cvRound(double)
+#endif
+
+#endif // OPENCV_CORE_CVDEF_H

3rdPart/include/opencv2/core/cvstd.inl.hpp

@@ -0,0 +1,286 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CVSTDINL_HPP
+#define OPENCV_CORE_CVSTDINL_HPP
+
+#include <complex>
+#include <ostream>
+#include <sstream>
+
+//! @cond IGNORED
+
+#ifdef _MSC_VER
+#pragma warning( push )
+#pragma warning( disable: 4127 )
+#endif
+
+namespace cv
+{
+
+template<typename _Tp> class DataType< std::complex<_Tp> >
+{
+public:
+    typedef std::complex<_Tp>  value_type;
+    typedef value_type         work_type;
+    typedef _Tp                channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels) };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+inline
+String::String(const std::string& str)
+    : cstr_(0), len_(0)
+{
+    size_t len = str.size();
+    if (len) memcpy(allocate(len), str.c_str(), len);
+}
+
+inline
+String::String(const std::string& str, size_t pos, size_t len)
+    : cstr_(0), len_(0)
+{
+    size_t strlen = str.size();
+    pos = min(pos, strlen);
+    len = min(strlen - pos, len);
+    if (!len) return;
+    memcpy(allocate(len), str.c_str() + pos, len);
+}
+
+inline
+String& String::operator = (const std::string& str)
+{
+    deallocate();
+    size_t len = str.size();
+    if (len) memcpy(allocate(len), str.c_str(), len);
+    return *this;
+}
+
+inline
+String& String::operator += (const std::string& str)
+{
+    *this = *this + str;
+    return *this;
+}
+
+inline
+String::operator std::string() const
+{
+    return std::string(cstr_, len_);
+}
+
+inline
+String operator + (const String& lhs, const std::string& rhs)
+{
+    String s;
+    size_t rhslen = rhs.size();
+    s.allocate(lhs.len_ + rhslen);
+    if (lhs.len_) memcpy(s.cstr_, lhs.cstr_, lhs.len_);
+    if (rhslen) memcpy(s.cstr_ + lhs.len_, rhs.c_str(), rhslen);
+    return s;
+}
+
+inline
+String operator + (const std::string& lhs, const String& rhs)
+{
+    String s;
+    size_t lhslen = lhs.size();
+    s.allocate(lhslen + rhs.len_);
+    if (lhslen) memcpy(s.cstr_, lhs.c_str(), lhslen);
+    if (rhs.len_) memcpy(s.cstr_ + lhslen, rhs.cstr_, rhs.len_);
+    return s;
+}
+
+inline
+FileNode::operator std::string() const
+{
+    String value;
+    read(*this, value, value);
+    return value;
+}
+
+template<> inline
+void operator >> (const FileNode& n, std::string& value)
+{
+    read(n, value, std::string());
+}
+
+template<> inline
+FileStorage& operator << (FileStorage& fs, const std::string& value)
+{
+    return fs << cv::String(value);
+}
+
+static inline
+std::ostream& operator << (std::ostream& os, const String& str)
+{
+    return os << str.c_str();
+}
+
+static inline
+std::ostream& operator << (std::ostream& out, Ptr<Formatted> fmtd)
+{
+    fmtd->reset();
+    for(const char* str = fmtd->next(); str; str = fmtd->next())
+        out << str;
+    return out;
+}
+
+static inline
+std::ostream& operator << (std::ostream& out, const Mat& mtx)
+{
+    return out << Formatter::get()->format(mtx);
+}
+
+static inline
+std::ostream& operator << (std::ostream& out, const UMat& m)
+{
+    return out << m.getMat(ACCESS_READ);
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Complex<_Tp>& c)
+{
+    return out << "(" << c.re << "," << c.im << ")";
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const std::vector<Point_<_Tp> >& vec)
+{
+    return out << Formatter::get()->format(Mat(vec));
+}
+
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const std::vector<Point3_<_Tp> >& vec)
+{
+    return out << Formatter::get()->format(Mat(vec));
+}
+
+
+template<typename _Tp, int m, int n> static inline
+std::ostream& operator << (std::ostream& out, const Matx<_Tp, m, n>& matx)
+{
+    return out << Formatter::get()->format(Mat(matx));
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Point_<_Tp>& p)
+{
+    out << "[" << p.x << ", " << p.y << "]";
+    return out;
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Point3_<_Tp>& p)
+{
+    out << "[" << p.x << ", " << p.y << ", " << p.z << "]";
+    return out;
+}
+
+template<typename _Tp, int n> static inline
+std::ostream& operator << (std::ostream& out, const Vec<_Tp, n>& vec)
+{
+    out << "[";
+    if (cv::traits::Depth<_Tp>::value <= CV_32S)
+    {
+        for (int i = 0; i < n - 1; ++i) {
+            out << (int)vec[i] << ", ";
+        }
+        out << (int)vec[n-1] << "]";
+    }
+    else
+    {
+        for (int i = 0; i < n - 1; ++i) {
+            out << vec[i] << ", ";
+        }
+        out << vec[n-1] << "]";
+    }
+
+    return out;
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Size_<_Tp>& size)
+{
+    return out << "[" << size.width << " x " << size.height << "]";
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Rect_<_Tp>& rect)
+{
+    return out << "[" << rect.width << " x " << rect.height << " from (" << rect.x << ", " << rect.y << ")]";
+}
+
+static inline std::ostream& operator << (std::ostream& out, const MatSize& msize)
+{
+    int i, dims = msize.dims();
+    for( i = 0; i < dims; i++ )
+    {
+        out << msize[i];
+        if( i < dims-1 )
+            out << " x ";
+    }
+    return out;
+}
+
+static inline std::ostream &operator<< (std::ostream &s, cv::Range &r)
+{
+    return s << "[" << r.start << " : " << r.end << ")";
+}
+
+} // cv
+
+#ifdef _MSC_VER
+#pragma warning( pop )
+#endif
+
+//! @endcond
+
+#endif // OPENCV_CORE_CVSTDINL_HPP

3rdPart/include/opencv2/core/detail/async_promise.hpp

@@ -0,0 +1,71 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_ASYNC_PROMISE_HPP
+#define OPENCV_CORE_ASYNC_PROMISE_HPP
+
+#include "../async.hpp"
+
+#include "exception_ptr.hpp"
+
+namespace cv {
+
+/** @addtogroup core_async
+@{
+*/
+
+
+/** @brief Provides result of asynchronous operations
+
+*/
+class CV_EXPORTS AsyncPromise
+{
+public:
+    ~AsyncPromise() CV_NOEXCEPT;
+    AsyncPromise() CV_NOEXCEPT;
+    explicit AsyncPromise(const AsyncPromise& o) CV_NOEXCEPT;
+    AsyncPromise& operator=(const AsyncPromise& o) CV_NOEXCEPT;
+    void release() CV_NOEXCEPT;
+
+    /** Returns associated AsyncArray
+    @note Can be called once
+    */
+    AsyncArray getArrayResult();
+
+    /** Stores asynchronous result.
+    @param[in] value result
+    */
+    void setValue(InputArray value);
+
+    // TODO "move" setters
+
+#if CV__EXCEPTION_PTR
+    /** Stores exception.
+    @param[in] exception exception to be raised in AsyncArray
+    */
+    void setException(std::exception_ptr exception);
+#endif
+
+    /** Stores exception.
+    @param[in] exception exception to be raised in AsyncArray
+    */
+    void setException(const cv::Exception& exception);
+
+#ifdef CV_CXX11
+    explicit AsyncPromise(AsyncPromise&& o) { p = o.p; o.p = NULL; }
+    AsyncPromise& operator=(AsyncPromise&& o) CV_NOEXCEPT { std::swap(p, o.p); return *this; }
+#endif
+
+
+    // PImpl
+    typedef struct AsyncArray::Impl Impl; friend struct AsyncArray::Impl;
+    inline void* _getImpl() const CV_NOEXCEPT { return p; }
+protected:
+    Impl* p;
+};
+
+
+//! @}
+} // namespace
+#endif // OPENCV_CORE_ASYNC_PROMISE_HPP

3rdPart/include/opencv2/core/detail/exception_ptr.hpp

@@ -0,0 +1,27 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_DETAILS_EXCEPTION_PTR_H
+#define OPENCV_CORE_DETAILS_EXCEPTION_PTR_H
+
+#ifndef CV__EXCEPTION_PTR
+#  if defined(__ANDROID__) && defined(ATOMIC_INT_LOCK_FREE) && ATOMIC_INT_LOCK_FREE < 2
+#    define CV__EXCEPTION_PTR 0  // Not supported, details: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58938
+#  elif defined(CV_CXX11)
+#    define CV__EXCEPTION_PTR 1
+#  elif defined(_MSC_VER)
+#    define CV__EXCEPTION_PTR (_MSC_VER >= 1600)
+#  elif defined(__clang__)
+#    define CV__EXCEPTION_PTR 0  // C++11 only (see above)
+#  elif defined(__GNUC__) && defined(__GXX_EXPERIMENTAL_CXX0X__)
+#    define CV__EXCEPTION_PTR (__GXX_EXPERIMENTAL_CXX0X__ > 0)
+#  endif
+#endif
+#ifndef CV__EXCEPTION_PTR
+#  define CV__EXCEPTION_PTR 0
+#elif CV__EXCEPTION_PTR
+#  include <exception>  // std::exception_ptr
+#endif
+
+#endif // OPENCV_CORE_DETAILS_EXCEPTION_PTR_H

3rdPart/include/opencv2/core/directx.hpp

@@ -0,0 +1,184 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the copyright holders or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_DIRECTX_HPP
+#define OPENCV_CORE_DIRECTX_HPP
+
+#include "mat.hpp"
+#include "ocl.hpp"
+
+#if !defined(__d3d11_h__)
+struct ID3D11Device;
+struct ID3D11Texture2D;
+#endif
+
+#if !defined(__d3d10_h__)
+struct ID3D10Device;
+struct ID3D10Texture2D;
+#endif
+
+#if !defined(_D3D9_H_)
+struct IDirect3DDevice9;
+struct IDirect3DDevice9Ex;
+struct IDirect3DSurface9;
+#endif
+
+
+namespace cv { namespace directx {
+
+namespace ocl {
+using namespace cv::ocl;
+
+//! @addtogroup core_directx
+// This section describes OpenCL and DirectX interoperability.
+//
+// To enable DirectX support, configure OpenCV using CMake with WITH_DIRECTX=ON . Note, DirectX is
+// supported only on Windows.
+//
+// To use OpenCL functionality you should first initialize OpenCL context from DirectX resource.
+//
+//! @{
+
+// TODO static functions in the Context class
+//! @brief Creates OpenCL context from D3D11 device
+//
+//! @param pD3D11Device - pointer to D3D11 device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromD3D11Device(ID3D11Device* pD3D11Device);
+
+//! @brief Creates OpenCL context from D3D10 device
+//
+//! @param pD3D10Device - pointer to D3D10 device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromD3D10Device(ID3D10Device* pD3D10Device);
+
+//! @brief Creates OpenCL context from Direct3DDevice9Ex device
+//
+//! @param pDirect3DDevice9Ex - pointer to Direct3DDevice9Ex device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromDirect3DDevice9Ex(IDirect3DDevice9Ex* pDirect3DDevice9Ex);
+
+//! @brief Creates OpenCL context from Direct3DDevice9 device
+//
+//! @param pDirect3DDevice9 - pointer to Direct3Device9 device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromDirect3DDevice9(IDirect3DDevice9* pDirect3DDevice9);
+
+//! @}
+
+} // namespace cv::directx::ocl
+
+//! @addtogroup core_directx
+//! @{
+
+//! @brief Converts InputArray to ID3D11Texture2D. If destination texture format is DXGI_FORMAT_NV12 then
+//!        input UMat expected to be in BGR format and data will be downsampled and color-converted to NV12.
+//
+//! @note Note: Destination texture must be allocated by application. Function does memory copy from src to
+//!             pD3D11Texture2D
+//
+//! @param src - source InputArray
+//! @param pD3D11Texture2D - destination D3D11 texture
+CV_EXPORTS void convertToD3D11Texture2D(InputArray src, ID3D11Texture2D* pD3D11Texture2D);
+
+//! @brief Converts ID3D11Texture2D to OutputArray. If input texture format is DXGI_FORMAT_NV12 then
+//!        data will be upsampled and color-converted to BGR format.
+//
+//! @note Note: Destination matrix will be re-allocated if it has not enough memory to match texture size.
+//!             function does memory copy from pD3D11Texture2D to dst
+//
+//! @param pD3D11Texture2D - source D3D11 texture
+//! @param dst             - destination OutputArray
+CV_EXPORTS void convertFromD3D11Texture2D(ID3D11Texture2D* pD3D11Texture2D, OutputArray dst);
+
+//! @brief Converts InputArray to ID3D10Texture2D
+//
+//! @note Note: function does memory copy from src to
+//!             pD3D10Texture2D
+//
+//! @param src             - source InputArray
+//! @param pD3D10Texture2D - destination D3D10 texture
+CV_EXPORTS void convertToD3D10Texture2D(InputArray src, ID3D10Texture2D* pD3D10Texture2D);
+
+//! @brief Converts ID3D10Texture2D to OutputArray
+//
+//! @note Note: function does memory copy from pD3D10Texture2D
+//!             to dst
+//
+//! @param pD3D10Texture2D - source D3D10 texture
+//! @param dst             - destination OutputArray
+CV_EXPORTS void convertFromD3D10Texture2D(ID3D10Texture2D* pD3D10Texture2D, OutputArray dst);
+
+//! @brief Converts InputArray to IDirect3DSurface9
+//
+//! @note Note: function does memory copy from src to
+//!             pDirect3DSurface9
+//
+//! @param src                 - source InputArray
+//! @param pDirect3DSurface9   - destination D3D10 texture
+//! @param surfaceSharedHandle - shared handle
+CV_EXPORTS void convertToDirect3DSurface9(InputArray src, IDirect3DSurface9* pDirect3DSurface9, void* surfaceSharedHandle = NULL);
+
+//! @brief Converts IDirect3DSurface9 to OutputArray
+//
+//! @note Note: function does memory copy from pDirect3DSurface9
+//!             to dst
+//
+//! @param pDirect3DSurface9   - source D3D10 texture
+//! @param dst                 - destination OutputArray
+//! @param surfaceSharedHandle - shared handle
+CV_EXPORTS void convertFromDirect3DSurface9(IDirect3DSurface9* pDirect3DSurface9, OutputArray dst, void* surfaceSharedHandle = NULL);
+
+//! @brief Get OpenCV type from DirectX type
+//! @param iDXGI_FORMAT - enum DXGI_FORMAT for D3D10/D3D11
+//! @return OpenCV type or -1 if there is no equivalent
+CV_EXPORTS int getTypeFromDXGI_FORMAT(const int iDXGI_FORMAT); // enum DXGI_FORMAT for D3D10/D3D11
+
+//! @brief Get OpenCV type from DirectX type
+//! @param iD3DFORMAT - enum D3DTYPE for D3D9
+//! @return OpenCV type or -1 if there is no equivalent
+CV_EXPORTS int getTypeFromD3DFORMAT(const int iD3DFORMAT); // enum D3DTYPE for D3D9
+
+//! @}
+
+} } // namespace cv::directx
+
+#endif // OPENCV_CORE_DIRECTX_HPP

3rdPart/include/opencv2/core/eigen.hpp

@@ -0,0 +1,402 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#ifndef OPENCV_CORE_EIGEN_HPP
+#define OPENCV_CORE_EIGEN_HPP
+
+#ifndef EIGEN_WORLD_VERSION
+#error "Wrong usage of OpenCV's Eigen utility header. Include Eigen's headers first. See https://github.com/opencv/opencv/issues/17366"
+#endif
+
+#include "opencv2/core.hpp"
+
+#if defined _MSC_VER && _MSC_VER >= 1200
+#define NOMINMAX // fix https://github.com/opencv/opencv/issues/17548
+#pragma warning( disable: 4714 ) //__forceinline is not inlined
+#pragma warning( disable: 4127 ) //conditional expression is constant
+#pragma warning( disable: 4244 ) //conversion from '__int64' to 'int', possible loss of data
+#endif
+
+#if !defined(OPENCV_DISABLE_EIGEN_TENSOR_SUPPORT)
+#if EIGEN_WORLD_VERSION == 3 && EIGEN_MAJOR_VERSION >= 3 \
+    && defined(CV_CXX11) && defined(CV_CXX_STD_ARRAY)
+#include <unsupported/Eigen/CXX11/Tensor>
+#define OPENCV_EIGEN_TENSOR_SUPPORT 1
+#endif  // EIGEN_WORLD_VERSION == 3 && EIGEN_MAJOR_VERSION >= 3
+#endif  // !defined(OPENCV_DISABLE_EIGEN_TENSOR_SUPPORT)
+
+namespace cv
+{
+
+/** @addtogroup core_eigen
+These functions are provided for OpenCV-Eigen interoperability. They convert `Mat`
+objects to corresponding `Eigen::Matrix` objects and vice-versa. Consult the [Eigen
+documentation](https://eigen.tuxfamily.org/dox/group__TutorialMatrixClass.html) for
+information about the `Matrix` template type.
+
+@note Using these functions requires the `Eigen/Dense` or similar header to be
+included before this header.
+*/
+//! @{
+
+#if defined(OPENCV_EIGEN_TENSOR_SUPPORT) || defined(CV_DOXYGEN)
+/** @brief Converts an Eigen::Tensor to a cv::Mat.
+
+The method converts an Eigen::Tensor with shape (H x W x C) to a cv::Mat where:
+ H = number of rows
+ W = number of columns
+ C = number of channels
+
+Usage:
+\code
+Eigen::Tensor<float, 3, Eigen::RowMajor> a_tensor(...);
+// populate tensor with values
+Mat a_mat;
+eigen2cv(a_tensor, a_mat);
+\endcode
+*/
+template <typename _Tp, int _layout> static inline
+void eigen2cv( const Eigen::Tensor<_Tp, 3, _layout> &src, OutputArray dst )
+{
+    if( !(_layout & Eigen::RowMajorBit) )
+    {
+        const std::array<int, 3> shuffle{2, 1, 0};
+        Eigen::Tensor<_Tp, 3, !_layout> row_major_tensor = src.swap_layout().shuffle(shuffle);
+        Mat _src(src.dimension(0), src.dimension(1), CV_MAKETYPE(DataType<_Tp>::type, src.dimension(2)), row_major_tensor.data());
+        _src.copyTo(dst);
+    }
+    else
+    {
+        Mat _src(src.dimension(0), src.dimension(1), CV_MAKETYPE(DataType<_Tp>::type, src.dimension(2)), (void *)src.data());
+        _src.copyTo(dst);
+    }
+}
+
+/** @brief Converts a cv::Mat to an Eigen::Tensor.
+
+The method converts a cv::Mat to an Eigen Tensor with shape (H x W x C) where:
+ H = number of rows
+ W = number of columns
+ C = number of channels
+
+Usage:
+\code
+Mat a_mat(...);
+// populate Mat with values
+Eigen::Tensor<float, 3, Eigen::RowMajor> a_tensor(...);
+cv2eigen(a_mat, a_tensor);
+\endcode
+*/
+template <typename _Tp, int _layout> static inline
+void cv2eigen( const Mat &src, Eigen::Tensor<_Tp, 3, _layout> &dst )
+{
+    if( !(_layout & Eigen::RowMajorBit) )
+    {
+        Eigen::Tensor<_Tp, 3, !_layout> row_major_tensor(src.rows, src.cols, src.channels());
+        Mat _dst(src.rows, src.cols, CV_MAKETYPE(DataType<_Tp>::type, src.channels()), row_major_tensor.data());
+        if (src.type() == _dst.type())
+            src.copyTo(_dst);
+        else
+            src.convertTo(_dst, _dst.type());
+        const std::array<int, 3> shuffle{2, 1, 0};
+        dst = row_major_tensor.swap_layout().shuffle(shuffle);
+    }
+    else
+    {
+        dst.resize(src.rows, src.cols, src.channels());
+        Mat _dst(src.rows, src.cols, CV_MAKETYPE(DataType<_Tp>::type, src.channels()), dst.data());
+        if (src.type() == _dst.type())
+            src.copyTo(_dst);
+        else
+            src.convertTo(_dst, _dst.type());
+    }
+}
+
+/** @brief Maps cv::Mat data to an Eigen::TensorMap.
+
+The method wraps an existing Mat data array with an Eigen TensorMap of shape (H x W x C) where:
+ H = number of rows
+ W = number of columns
+ C = number of channels
+
+Explicit instantiation of the return type is required.
+
+@note Caller should be aware of the lifetime of the cv::Mat instance and take appropriate safety measures.
+The cv::Mat instance will retain ownership of the data and the Eigen::TensorMap will lose access when the cv::Mat data is deallocated.
+
+The example below initializes a cv::Mat and produces an Eigen::TensorMap:
+\code
+float arr[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+Mat a_mat(2, 2, CV_32FC3, arr);
+Eigen::TensorMap<Eigen::Tensor<float, 3, Eigen::RowMajor>> a_tensormap = cv2eigen_tensormap<float>(a_mat);
+\endcode
+*/
+template <typename _Tp> static inline
+Eigen::TensorMap<Eigen::Tensor<_Tp, 3, Eigen::RowMajor>> cv2eigen_tensormap(InputArray src)
+{
+    Mat mat = src.getMat();
+    CV_CheckTypeEQ(mat.type(), CV_MAKETYPE(traits::Type<_Tp>::value, mat.channels()), "");
+    return Eigen::TensorMap<Eigen::Tensor<_Tp, 3, Eigen::RowMajor>>((_Tp *)mat.data, mat.rows, mat.cols, mat.channels());
+}
+#endif // OPENCV_EIGEN_TENSOR_SUPPORT
+
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& src, OutputArray dst )
+{
+    if( !(src.Flags & Eigen::RowMajorBit) )
+    {
+        Mat _src(src.cols(), src.rows(), traits::Type<_Tp>::value,
+              (void*)src.data(), src.outerStride()*sizeof(_Tp));
+        transpose(_src, dst);
+    }
+    else
+    {
+        Mat _src(src.rows(), src.cols(), traits::Type<_Tp>::value,
+                 (void*)src.data(), src.outerStride()*sizeof(_Tp));
+        _src.copyTo(dst);
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& src,
+               Matx<_Tp, _rows, _cols>& dst )
+{
+    if( !(src.Flags & Eigen::RowMajorBit) )
+    {
+        dst = Matx<_Tp, _cols, _rows>(static_cast<const _Tp*>(src.data())).t();
+    }
+    else
+    {
+        dst = Matx<_Tp, _rows, _cols>(static_cast<const _Tp*>(src.data()));
+    }
+}
+
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& dst )
+{
+    CV_DbgAssert(src.rows == _rows && src.cols == _cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else if( src.cols == src.rows )
+        {
+            src.convertTo(_dst, _dst.type());
+            transpose(_dst, _dst);
+        }
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void cv2eigen( const Matx<_Tp, _rows, _cols>& src,
+               Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& dst )
+{
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(_cols, _rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(_rows, _cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        Mat(src).copyTo(_dst);
+    }
+}
+
+template<typename _Tp>  static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, Eigen::Dynamic>& dst )
+{
+    dst.resize(src.rows, src.cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+             dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else if( src.cols == src.rows )
+        {
+            src.convertTo(_dst, _dst.type());
+            transpose(_dst, _dst);
+        }
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows, int _cols> static inline
+void cv2eigen( const Matx<_Tp, _rows, _cols>& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, Eigen::Dynamic>& dst )
+{
+    dst.resize(_rows, _cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(_cols, _rows, traits::Type<_Tp>::value,
+             dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(_rows, _cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        Mat(src).copyTo(_dst);
+    }
+}
+
+template<typename _Tp> static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, 1>& dst )
+{
+    CV_Assert(src.cols == 1);
+    dst.resize(src.rows);
+
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows> static inline
+void cv2eigen( const Matx<_Tp, _rows, 1>& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, 1>& dst )
+{
+    dst.resize(_rows);
+
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(1, _rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(_rows, 1, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.copyTo(_dst);
+    }
+}
+
+
+template<typename _Tp> static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, 1, Eigen::Dynamic>& dst )
+{
+    CV_Assert(src.rows == 1);
+    dst.resize(src.cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+//Matx
+template<typename _Tp, int _cols> static inline
+void cv2eigen( const Matx<_Tp, 1, _cols>& src,
+               Eigen::Matrix<_Tp, 1, Eigen::Dynamic>& dst )
+{
+    dst.resize(_cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(_cols, 1, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(1, _cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        Mat(src).copyTo(_dst);
+    }
+}
+
+//! @}
+
+} // cv
+
+#endif

3rdPart/include/opencv2/core/fast_math.hpp

@@ -0,0 +1,411 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_FAST_MATH_HPP
+#define OPENCV_CORE_FAST_MATH_HPP
+
+#include "opencv2/core/cvdef.h"
+
+//! @addtogroup core_utils
+//! @{
+
+/****************************************************************************************\
+*                                      fast math                                         *
+\****************************************************************************************/
+
+#ifdef __cplusplus
+#  include <cmath>
+#else
+#  ifdef __BORLANDC__
+#    include <fastmath.h>
+#  else
+#    include <math.h>
+#  endif
+#endif
+
+#if defined(__CUDACC__)
+  // nothing, intrinsics/asm code is not supported
+#else
+  #if ((defined _MSC_VER && defined _M_X64) \
+      || (defined __GNUC__ && defined __x86_64__ && defined __SSE2__)) \
+      && !defined(OPENCV_SKIP_INCLUDE_EMMINTRIN_H)
+    #include <emmintrin.h>
+  #endif
+
+  #if defined __PPC64__ && defined __GNUC__ && defined _ARCH_PWR8 \
+      && !defined(OPENCV_SKIP_INCLUDE_ALTIVEC_H)
+    #include <altivec.h>
+    #undef vector
+    #undef bool
+    #undef pixel
+  #endif
+
+  #if defined(CV_INLINE_ROUND_FLT)
+    // user-specified version
+    // CV_INLINE_ROUND_DBL should be defined too
+  #elif defined __GNUC__ && defined __arm__ && (defined __ARM_PCS_VFP || defined __ARM_VFPV3__ || defined __ARM_NEON__) && !defined __SOFTFP__
+    // 1. general scheme
+    #define ARM_ROUND(_value, _asm_string) \
+        int res; \
+        float temp; \
+        CV_UNUSED(temp); \
+        __asm__(_asm_string : [res] "=r" (res), [temp] "=w" (temp) : [value] "w" (_value)); \
+        return res
+    // 2. version for double
+    #ifdef __clang__
+        #define CV_INLINE_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %[value] \n vmov %[res], %[temp]")
+    #else
+        #define CV_INLINE_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %P[value] \n vmov %[res], %[temp]")
+    #endif
+    // 3. version for float
+    #define CV_INLINE_ROUND_FLT(value) ARM_ROUND(value, "vcvtr.s32.f32 %[temp], %[value]\n vmov %[res], %[temp]")
+  #elif defined __PPC64__ && defined __GNUC__ && defined _ARCH_PWR8
+    // P8 and newer machines can convert fp32/64 to int quickly.
+    #define CV_INLINE_ROUND_DBL(value) \
+        int out; \
+        double temp; \
+        __asm__( "fctiw %[temp],%[in]\n\tmfvsrwz %[out],%[temp]\n\t" : [out] "=r" (out), [temp] "=d" (temp) : [in] "d" ((double)(value)) : ); \
+        return out;
+
+    // FP32 also works with FP64 routine above
+    #define CV_INLINE_ROUND_FLT(value) CV_INLINE_ROUND_DBL(value)
+  #endif
+
+  #ifdef CV_INLINE_ISINF_FLT
+    // user-specified version
+    // CV_INLINE_ISINF_DBL should be defined too
+  #elif defined __PPC64__ && defined _ARCH_PWR9 && defined(scalar_test_data_class)
+    #define CV_INLINE_ISINF_DBL(value) return scalar_test_data_class(value, 0x30);
+    #define CV_INLINE_ISINF_FLT(value) CV_INLINE_ISINF_DBL(value)
+  #endif
+
+  #ifdef CV_INLINE_ISNAN_FLT
+    // user-specified version
+    // CV_INLINE_ISNAN_DBL should be defined too
+  #elif defined __PPC64__ && defined _ARCH_PWR9 && defined(scalar_test_data_class)
+    #define CV_INLINE_ISNAN_DBL(value) return scalar_test_data_class(value, 0x40);
+    #define CV_INLINE_ISNAN_FLT(value) CV_INLINE_ISNAN_DBL(value)
+  #endif
+
+  #if !defined(OPENCV_USE_FASTMATH_BUILTINS) \
+    && ( \
+        defined(__x86_64__) || defined(__i686__) \
+        || defined(__arm__) \
+        || defined(__PPC64__) \
+    )
+    /* Let builtin C math functions when available. Dedicated hardware is available to
+       round and convert FP values. */
+    #define OPENCV_USE_FASTMATH_BUILTINS 1
+  #endif
+
+  /* Enable builtin math functions if possible, desired, and available.
+     Note, not all math functions inline equally. E.g lrint will not inline
+     without the -fno-math-errno option. */
+  #if defined(CV_ICC)
+    // nothing
+  #elif defined(OPENCV_USE_FASTMATH_BUILTINS) && OPENCV_USE_FASTMATH_BUILTINS
+    #if defined(__clang__)
+      #define CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS
+      #if !defined(CV_INLINE_ISNAN_DBL) && __has_builtin(__builtin_isnan)
+        #define CV_INLINE_ISNAN_DBL(value) return __builtin_isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISNAN_FLT) && __has_builtin(__builtin_isnan)
+        #define CV_INLINE_ISNAN_FLT(value) return __builtin_isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_DBL) && __has_builtin(__builtin_isinf)
+        #define CV_INLINE_ISINF_DBL(value) return __builtin_isinf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_FLT) && __has_builtin(__builtin_isinf)
+        #define CV_INLINE_ISINF_FLT(value) return __builtin_isinf(value);
+      #endif
+    #elif defined(__GNUC__)
+      #define CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS
+      #if !defined(CV_INLINE_ISNAN_DBL)
+        #define CV_INLINE_ISNAN_DBL(value) return __builtin_isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISNAN_FLT)
+        #define CV_INLINE_ISNAN_FLT(value) return __builtin_isnanf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_DBL)
+        #define CV_INLINE_ISINF_DBL(value) return __builtin_isinf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_FLT)
+        #define CV_INLINE_ISINF_FLT(value) return __builtin_isinff(value);
+      #endif
+    #elif defined(_MSC_VER)
+      #if !defined(CV_INLINE_ISNAN_DBL)
+        #define CV_INLINE_ISNAN_DBL(value) return isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISNAN_FLT)
+        #define CV_INLINE_ISNAN_FLT(value) return isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_DBL)
+        #define CV_INLINE_ISINF_DBL(value) return isinf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_FLT)
+        #define CV_INLINE_ISINF_FLT(value) return isinf(value);
+      #endif
+    #endif
+  #endif
+
+#endif // defined(__CUDACC__)
+
+/** @brief Rounds floating-point number to the nearest integer
+
+ @param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
+ result is not defined.
+ */
+CV_INLINE int
+cvRound( double value )
+{
+#if defined CV_INLINE_ROUND_DBL
+    CV_INLINE_ROUND_DBL(value);
+#elif ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ \
+    && defined __SSE2__ && !defined __APPLE__) || CV_SSE2) \
+    && !defined(__CUDACC__)
+    __m128d t = _mm_set_sd( value );
+    return _mm_cvtsd_si32(t);
+#elif defined _MSC_VER && defined _M_IX86
+    int t;
+    __asm
+    {
+        fld value;
+        fistp t;
+    }
+    return t;
+#elif defined CV_ICC || defined __GNUC__
+    return (int)(lrint(value));
+#else
+    /* it's ok if round does not comply with IEEE754 standard;
+       the tests should allow +/-1 difference when the tested functions use round */
+    return (int)(value + (value >= 0 ? 0.5 : -0.5));
+#endif
+}
+
+
+/** @brief Rounds floating-point number to the nearest integer not larger than the original.
+
+ The function computes an integer i such that:
+ \f[i \le \texttt{value} < i+1\f]
+ @param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
+ result is not defined.
+ */
+CV_INLINE int cvFloor( double value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_floor(value);
+#else
+    int i = (int)value;
+    return i - (i > value);
+#endif
+}
+
+/** @brief Rounds floating-point number to the nearest integer not smaller than the original.
+
+ The function computes an integer i such that:
+ \f[i \le \texttt{value} < i+1\f]
+ @param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
+ result is not defined.
+ */
+CV_INLINE int cvCeil( double value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_ceil(value);
+#else
+    int i = (int)value;
+    return i + (i < value);
+#endif
+}
+
+/** @brief Determines if the argument is Not A Number.
+
+ @param value The input floating-point value
+
+ The function returns 1 if the argument is Not A Number (as defined by IEEE754 standard), 0
+ otherwise. */
+CV_INLINE int cvIsNaN( double value )
+{
+#if defined CV_INLINE_ISNAN_DBL
+    CV_INLINE_ISNAN_DBL(value);
+#else
+    Cv64suf ieee754;
+    ieee754.f = value;
+    return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) +
+           ((unsigned)ieee754.u != 0) > 0x7ff00000;
+#endif
+}
+
+/** @brief Determines if the argument is Infinity.
+
+ @param value The input floating-point value
+
+ The function returns 1 if the argument is a plus or minus infinity (as defined by IEEE754 standard)
+ and 0 otherwise. */
+CV_INLINE int cvIsInf( double value )
+{
+#if defined CV_INLINE_ISINF_DBL
+    CV_INLINE_ISINF_DBL(value);
+#elif defined(__x86_64__) || defined(_M_X64) || defined(__aarch64__) || defined(_M_ARM64) || defined(__PPC64__)
+    Cv64suf ieee754;
+    ieee754.f = value;
+    return (ieee754.u & 0x7fffffff00000000) ==
+                        0x7ff0000000000000;
+#else
+    Cv64suf ieee754;
+    ieee754.f = value;
+    return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) == 0x7ff00000 &&
+            (unsigned)ieee754.u == 0;
+#endif
+}
+
+#ifdef __cplusplus
+
+/** @overload */
+CV_INLINE int cvRound(float value)
+{
+#if defined CV_INLINE_ROUND_FLT
+    CV_INLINE_ROUND_FLT(value);
+#elif ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ \
+    && defined __SSE2__ && !defined __APPLE__) || CV_SSE2) \
+    && !defined(__CUDACC__)
+    __m128 t = _mm_set_ss( value );
+    return _mm_cvtss_si32(t);
+#elif defined _MSC_VER && defined _M_IX86
+    int t;
+    __asm
+    {
+        fld value;
+        fistp t;
+    }
+    return t;
+#elif defined CV_ICC || defined __GNUC__
+    return (int)(lrintf(value));
+#else
+    /* it's ok if round does not comply with IEEE754 standard;
+     the tests should allow +/-1 difference when the tested functions use round */
+    return (int)(value + (value >= 0 ? 0.5f : -0.5f));
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvRound( int value )
+{
+    return value;
+}
+
+/** @overload */
+CV_INLINE int cvFloor( float value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_floorf(value);
+#else
+    int i = (int)value;
+    return i - (i > value);
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvFloor( int value )
+{
+    return value;
+}
+
+/** @overload */
+CV_INLINE int cvCeil( float value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_ceilf(value);
+#else
+    int i = (int)value;
+    return i + (i < value);
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvCeil( int value )
+{
+    return value;
+}
+
+/** @overload */
+CV_INLINE int cvIsNaN( float value )
+{
+#if defined CV_INLINE_ISNAN_FLT
+    CV_INLINE_ISNAN_FLT(value);
+#else
+    Cv32suf ieee754;
+    ieee754.f = value;
+    return (ieee754.u & 0x7fffffff) > 0x7f800000;
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvIsInf( float value )
+{
+#if defined CV_INLINE_ISINF_FLT
+    CV_INLINE_ISINF_FLT(value);
+#else
+    Cv32suf ieee754;
+    ieee754.f = value;
+    return (ieee754.u & 0x7fffffff) == 0x7f800000;
+#endif
+}
+
+#endif // __cplusplus
+
+//! @} core_utils
+
+#endif

3rdPart/include/opencv2/core/hal/hal.hpp

@@ -0,0 +1,250 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_HAL_HPP
+#define OPENCV_HAL_HPP
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/cvstd.hpp"
+#include "opencv2/core/hal/interface.h"
+
+namespace cv { namespace hal {
+
+//! @addtogroup core_hal_functions
+//! @{
+
+CV_EXPORTS int normHamming(const uchar* a, int n);
+CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n);
+
+CV_EXPORTS int normHamming(const uchar* a, int n, int cellSize);
+CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n, int cellSize);
+
+CV_EXPORTS int LU32f(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS int LU64f(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky32f(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky64f(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+CV_EXPORTS void SVD32f(float* At, size_t astep, float* W, float* U, size_t ustep, float* Vt, size_t vstep, int m, int n, int flags);
+CV_EXPORTS void SVD64f(double* At, size_t astep, double* W, double* U, size_t ustep, double* Vt, size_t vstep, int m, int n, int flags);
+CV_EXPORTS int QR32f(float* A, size_t astep, int m, int n, int k, float* b, size_t bstep, float* hFactors);
+CV_EXPORTS int QR64f(double* A, size_t astep, int m, int n, int k, double* b, size_t bstep, double* hFactors);
+
+CV_EXPORTS void gemm32f(const float* src1, size_t src1_step, const float* src2, size_t src2_step,
+                        float alpha, const float* src3, size_t src3_step, float beta, float* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+CV_EXPORTS void gemm64f(const double* src1, size_t src1_step, const double* src2, size_t src2_step,
+                        double alpha, const double* src3, size_t src3_step, double beta, double* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+CV_EXPORTS void gemm32fc(const float* src1, size_t src1_step, const float* src2, size_t src2_step,
+                        float alpha, const float* src3, size_t src3_step, float beta, float* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+CV_EXPORTS void gemm64fc(const double* src1, size_t src1_step, const double* src2, size_t src2_step,
+                        double alpha, const double* src3, size_t src3_step, double beta, double* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+
+CV_EXPORTS int normL1_(const uchar* a, const uchar* b, int n);
+CV_EXPORTS float normL1_(const float* a, const float* b, int n);
+CV_EXPORTS float normL2Sqr_(const float* a, const float* b, int n);
+
+CV_EXPORTS void exp32f(const float* src, float* dst, int n);
+CV_EXPORTS void exp64f(const double* src, double* dst, int n);
+CV_EXPORTS void log32f(const float* src, float* dst, int n);
+CV_EXPORTS void log64f(const double* src, double* dst, int n);
+
+CV_EXPORTS void fastAtan32f(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
+CV_EXPORTS void fastAtan64f(const double* y, const double* x, double* dst, int n, bool angleInDegrees);
+CV_EXPORTS void magnitude32f(const float* x, const float* y, float* dst, int n);
+CV_EXPORTS void magnitude64f(const double* x, const double* y, double* dst, int n);
+CV_EXPORTS void sqrt32f(const float* src, float* dst, int len);
+CV_EXPORTS void sqrt64f(const double* src, double* dst, int len);
+CV_EXPORTS void invSqrt32f(const float* src, float* dst, int len);
+CV_EXPORTS void invSqrt64f(const double* src, double* dst, int len);
+
+CV_EXPORTS void split8u(const uchar* src, uchar** dst, int len, int cn );
+CV_EXPORTS void split16u(const ushort* src, ushort** dst, int len, int cn );
+CV_EXPORTS void split32s(const int* src, int** dst, int len, int cn );
+CV_EXPORTS void split64s(const int64* src, int64** dst, int len, int cn );
+
+CV_EXPORTS void merge8u(const uchar** src, uchar* dst, int len, int cn );
+CV_EXPORTS void merge16u(const ushort** src, ushort* dst, int len, int cn );
+CV_EXPORTS void merge32s(const int** src, int* dst, int len, int cn );
+CV_EXPORTS void merge64s(const int64** src, int64* dst, int len, int cn );
+
+CV_EXPORTS void add8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void sub8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void max8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void min8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void absdiff8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void and8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void or8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void xor8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void not8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void cmp8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp8s(const schar* src1, size_t step1, const schar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp16s(const short* src1, size_t step1, const short* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp32s(const int* src1, size_t step1, const int* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp32f(const float* src1, size_t step1, const float* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp64f(const double* src1, size_t step1, const double* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+
+CV_EXPORTS void mul8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* scale);
+
+CV_EXPORTS void div8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* scale);
+
+CV_EXPORTS void recip8u( const uchar *, size_t, const uchar * src2, size_t step2, uchar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip8s( const schar *, size_t, const schar * src2, size_t step2, schar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip16u( const ushort *, size_t, const ushort * src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip16s( const short *, size_t, const short * src2, size_t step2, short* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip32s( const int *, size_t, const int * src2, size_t step2, int* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip32f( const float *, size_t, const float * src2, size_t step2, float* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip64f( const double *, size_t, const double * src2, size_t step2, double* dst, size_t step, int width, int height, void* scale);
+
+CV_EXPORTS void addWeighted8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _scalars );
+CV_EXPORTS void addWeighted8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* scalars );
+
+struct CV_EXPORTS DFT1D
+{
+    static Ptr<DFT1D> create(int len, int count, int depth, int flags, bool * useBuffer = 0);
+    virtual void apply(const uchar *src, uchar *dst) = 0;
+    virtual ~DFT1D() {}
+};
+
+struct CV_EXPORTS DFT2D
+{
+    static Ptr<DFT2D> create(int width, int height, int depth,
+                             int src_channels, int dst_channels,
+                             int flags, int nonzero_rows = 0);
+    virtual void apply(const uchar *src_data, size_t src_step, uchar *dst_data, size_t dst_step) = 0;
+    virtual ~DFT2D() {}
+};
+
+struct CV_EXPORTS DCT2D
+{
+    static Ptr<DCT2D> create(int width, int height, int depth, int flags);
+    virtual void apply(const uchar *src_data, size_t src_step, uchar *dst_data, size_t dst_step) = 0;
+    virtual ~DCT2D() {}
+};
+
+//! @} core_hal
+
+//=============================================================================
+// for binary compatibility with 3.0
+
+//! @cond IGNORED
+
+CV_EXPORTS int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+
+CV_EXPORTS void exp(const float* src, float* dst, int n);
+CV_EXPORTS void exp(const double* src, double* dst, int n);
+CV_EXPORTS void log(const float* src, float* dst, int n);
+CV_EXPORTS void log(const double* src, double* dst, int n);
+
+CV_EXPORTS void fastAtan2(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
+CV_EXPORTS void magnitude(const float* x, const float* y, float* dst, int n);
+CV_EXPORTS void magnitude(const double* x, const double* y, double* dst, int n);
+CV_EXPORTS void sqrt(const float* src, float* dst, int len);
+CV_EXPORTS void sqrt(const double* src, double* dst, int len);
+CV_EXPORTS void invSqrt(const float* src, float* dst, int len);
+CV_EXPORTS void invSqrt(const double* src, double* dst, int len);
+
+//! @endcond
+
+}} //cv::hal
+
+#endif //OPENCV_HAL_HPP

3rdPart/include/opencv2/core/hal/interface.h

@@ -0,0 +1,182 @@
+#ifndef OPENCV_CORE_HAL_INTERFACE_H
+#define OPENCV_CORE_HAL_INTERFACE_H
+
+//! @addtogroup core_hal_interface
+//! @{
+
+//! @name Return codes
+//! @{
+#define CV_HAL_ERROR_OK 0
+#define CV_HAL_ERROR_NOT_IMPLEMENTED 1
+#define CV_HAL_ERROR_UNKNOWN -1
+//! @}
+
+#ifdef __cplusplus
+#include <cstddef>
+#else
+#include <stddef.h>
+#include <stdbool.h>
+#endif
+
+//! @name Data types
+//! primitive types
+//! - schar  - signed 1 byte integer
+//! - uchar  - unsigned 1 byte integer
+//! - short  - signed 2 byte integer
+//! - ushort - unsigned 2 byte integer
+//! - int    - signed 4 byte integer
+//! - uint   - unsigned 4 byte integer
+//! - int64  - signed 8 byte integer
+//! - uint64 - unsigned 8 byte integer
+//! @{
+#if !defined _MSC_VER && !defined __BORLANDC__
+#  if defined __cplusplus && __cplusplus >= 201103L && !defined __APPLE__
+#    include <cstdint>
+#    ifdef __NEWLIB__
+        typedef unsigned int uint;
+#    else
+        typedef std::uint32_t uint;
+#    endif
+#  else
+#    include <stdint.h>
+     typedef uint32_t uint;
+#  endif
+#else
+   typedef unsigned uint;
+#endif
+
+typedef signed char schar;
+
+#ifndef __IPL_H__
+   typedef unsigned char uchar;
+   typedef unsigned short ushort;
+#endif
+
+#if defined _MSC_VER || defined __BORLANDC__
+   typedef __int64 int64;
+   typedef unsigned __int64 uint64;
+#  define CV_BIG_INT(n)   n##I64
+#  define CV_BIG_UINT(n)  n##UI64
+#else
+   typedef int64_t int64;
+   typedef uint64_t uint64;
+#  define CV_BIG_INT(n)   n##LL
+#  define CV_BIG_UINT(n)  n##ULL
+#endif
+
+#define CV_CN_MAX     512
+#define CV_CN_SHIFT   3
+#define CV_DEPTH_MAX  (1 << CV_CN_SHIFT)
+
+#define CV_8U   0
+#define CV_8S   1
+#define CV_16U  2
+#define CV_16S  3
+#define CV_32S  4
+#define CV_32F  5
+#define CV_64F  6
+#define CV_USRTYPE1 7
+
+#define CV_MAT_DEPTH_MASK       (CV_DEPTH_MAX - 1)
+#define CV_MAT_DEPTH(flags)     ((flags) & CV_MAT_DEPTH_MASK)
+
+#define CV_MAKETYPE(depth,cn) (CV_MAT_DEPTH(depth) + (((cn)-1) << CV_CN_SHIFT))
+#define CV_MAKE_TYPE CV_MAKETYPE
+
+#define CV_8UC1 CV_MAKETYPE(CV_8U,1)
+#define CV_8UC2 CV_MAKETYPE(CV_8U,2)
+#define CV_8UC3 CV_MAKETYPE(CV_8U,3)
+#define CV_8UC4 CV_MAKETYPE(CV_8U,4)
+#define CV_8UC(n) CV_MAKETYPE(CV_8U,(n))
+
+#define CV_8SC1 CV_MAKETYPE(CV_8S,1)
+#define CV_8SC2 CV_MAKETYPE(CV_8S,2)
+#define CV_8SC3 CV_MAKETYPE(CV_8S,3)
+#define CV_8SC4 CV_MAKETYPE(CV_8S,4)
+#define CV_8SC(n) CV_MAKETYPE(CV_8S,(n))
+
+#define CV_16UC1 CV_MAKETYPE(CV_16U,1)
+#define CV_16UC2 CV_MAKETYPE(CV_16U,2)
+#define CV_16UC3 CV_MAKETYPE(CV_16U,3)
+#define CV_16UC4 CV_MAKETYPE(CV_16U,4)
+#define CV_16UC(n) CV_MAKETYPE(CV_16U,(n))
+
+#define CV_16SC1 CV_MAKETYPE(CV_16S,1)
+#define CV_16SC2 CV_MAKETYPE(CV_16S,2)
+#define CV_16SC3 CV_MAKETYPE(CV_16S,3)
+#define CV_16SC4 CV_MAKETYPE(CV_16S,4)
+#define CV_16SC(n) CV_MAKETYPE(CV_16S,(n))
+
+#define CV_32SC1 CV_MAKETYPE(CV_32S,1)
+#define CV_32SC2 CV_MAKETYPE(CV_32S,2)
+#define CV_32SC3 CV_MAKETYPE(CV_32S,3)
+#define CV_32SC4 CV_MAKETYPE(CV_32S,4)
+#define CV_32SC(n) CV_MAKETYPE(CV_32S,(n))
+
+#define CV_32FC1 CV_MAKETYPE(CV_32F,1)
+#define CV_32FC2 CV_MAKETYPE(CV_32F,2)
+#define CV_32FC3 CV_MAKETYPE(CV_32F,3)
+#define CV_32FC4 CV_MAKETYPE(CV_32F,4)
+#define CV_32FC(n) CV_MAKETYPE(CV_32F,(n))
+
+#define CV_64FC1 CV_MAKETYPE(CV_64F,1)
+#define CV_64FC2 CV_MAKETYPE(CV_64F,2)
+#define CV_64FC3 CV_MAKETYPE(CV_64F,3)
+#define CV_64FC4 CV_MAKETYPE(CV_64F,4)
+#define CV_64FC(n) CV_MAKETYPE(CV_64F,(n))
+//! @}
+
+//! @name Comparison operation
+//! @sa cv::CmpTypes
+//! @{
+#define CV_HAL_CMP_EQ 0
+#define CV_HAL_CMP_GT 1
+#define CV_HAL_CMP_GE 2
+#define CV_HAL_CMP_LT 3
+#define CV_HAL_CMP_LE 4
+#define CV_HAL_CMP_NE 5
+//! @}
+
+//! @name Border processing modes
+//! @sa cv::BorderTypes
+//! @{
+#define CV_HAL_BORDER_CONSTANT 0
+#define CV_HAL_BORDER_REPLICATE 1
+#define CV_HAL_BORDER_REFLECT 2
+#define CV_HAL_BORDER_WRAP 3
+#define CV_HAL_BORDER_REFLECT_101 4
+#define CV_HAL_BORDER_TRANSPARENT 5
+#define CV_HAL_BORDER_ISOLATED 16
+//! @}
+
+//! @name DFT flags
+//! @{
+#define CV_HAL_DFT_INVERSE        1
+#define CV_HAL_DFT_SCALE          2
+#define CV_HAL_DFT_ROWS           4
+#define CV_HAL_DFT_COMPLEX_OUTPUT 16
+#define CV_HAL_DFT_REAL_OUTPUT    32
+#define CV_HAL_DFT_TWO_STAGE      64
+#define CV_HAL_DFT_STAGE_COLS    128
+#define CV_HAL_DFT_IS_CONTINUOUS 512
+#define CV_HAL_DFT_IS_INPLACE 1024
+//! @}
+
+//! @name SVD flags
+//! @{
+#define CV_HAL_SVD_NO_UV    1
+#define CV_HAL_SVD_SHORT_UV 2
+#define CV_HAL_SVD_MODIFY_A 4
+#define CV_HAL_SVD_FULL_UV  8
+//! @}
+
+//! @name Gemm flags
+//! @{
+#define CV_HAL_GEMM_1_T 1
+#define CV_HAL_GEMM_2_T 2
+#define CV_HAL_GEMM_3_T 4
+//! @}
+
+//! @}
+
+#endif

3rdPart/include/opencv2/core/hal/intrin.hpp

@@ -0,0 +1,698 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_HAL_INTRIN_HPP
+#define OPENCV_HAL_INTRIN_HPP
+
+#include <cmath>
+#include <float.h>
+#include <stdlib.h>
+#include "opencv2/core/cvdef.h"
+
+#define OPENCV_HAL_ADD(a, b) ((a) + (b))
+#define OPENCV_HAL_AND(a, b) ((a) & (b))
+#define OPENCV_HAL_NOP(a) (a)
+#define OPENCV_HAL_1ST(a, b) (a)
+
+namespace {
+inline unsigned int trailingZeros32(unsigned int value) {
+#if defined(_MSC_VER)
+#if (_MSC_VER < 1700) || defined(_M_ARM) || defined(_M_ARM64)
+    unsigned long index = 0;
+    _BitScanForward(&index, value);
+    return (unsigned int)index;
+#elif defined(__clang__)
+    // clang-cl doesn't export _tzcnt_u32 for non BMI systems
+    return value ? __builtin_ctz(value) : 32;
+#else
+    return _tzcnt_u32(value);
+#endif
+#elif defined(__GNUC__) || defined(__GNUG__)
+    return __builtin_ctz(value);
+#elif defined(__ICC) || defined(__INTEL_COMPILER)
+    return _bit_scan_forward(value);
+#elif defined(__clang__)
+    return llvm.cttz.i32(value, true);
+#else
+    static const int MultiplyDeBruijnBitPosition[32] = {
+        0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8,
+        31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9 };
+    return MultiplyDeBruijnBitPosition[((uint32_t)((value & -value) * 0x077CB531U)) >> 27];
+#endif
+}
+}
+
+// unlike HAL API, which is in cv::hal,
+// we put intrinsics into cv namespace to make its
+// access from within opencv code more accessible
+namespace cv {
+
+namespace hal {
+
+enum StoreMode
+{
+    STORE_UNALIGNED = 0,
+    STORE_ALIGNED = 1,
+    STORE_ALIGNED_NOCACHE = 2
+};
+
+}
+
+// TODO FIXIT: Don't use "God" traits. Split on separate cases.
+template<typename _Tp> struct V_TypeTraits
+{
+};
+
+#define CV_INTRIN_DEF_TYPE_TRAITS(type, int_type_, uint_type_, abs_type_, w_type_, q_type_, sum_type_) \
+    template<> struct V_TypeTraits<type> \
+    { \
+        typedef type value_type; \
+        typedef int_type_ int_type; \
+        typedef abs_type_ abs_type; \
+        typedef uint_type_ uint_type; \
+        typedef w_type_ w_type; \
+        typedef q_type_ q_type; \
+        typedef sum_type_ sum_type; \
+    \
+        static inline int_type reinterpret_int(type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.l = x; \
+            return v.i; \
+        } \
+    \
+        static inline type reinterpret_from_int(int_type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.i = x; \
+            return v.l; \
+        } \
+    }
+
+#define CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(type, int_type_, uint_type_, abs_type_, w_type_, sum_type_) \
+    template<> struct V_TypeTraits<type> \
+    { \
+        typedef type value_type; \
+        typedef int_type_ int_type; \
+        typedef abs_type_ abs_type; \
+        typedef uint_type_ uint_type; \
+        typedef w_type_ w_type; \
+        typedef sum_type_ sum_type; \
+    \
+        static inline int_type reinterpret_int(type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.l = x; \
+            return v.i; \
+        } \
+    \
+        static inline type reinterpret_from_int(int_type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.i = x; \
+            return v.l; \
+        } \
+    }
+
+CV_INTRIN_DEF_TYPE_TRAITS(uchar, schar, uchar, uchar, ushort, unsigned, unsigned);
+CV_INTRIN_DEF_TYPE_TRAITS(schar, schar, uchar, uchar, short, int, int);
+CV_INTRIN_DEF_TYPE_TRAITS(ushort, short, ushort, ushort, unsigned, uint64, unsigned);
+CV_INTRIN_DEF_TYPE_TRAITS(short, short, ushort, ushort, int, int64, int);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(unsigned, int, unsigned, unsigned, uint64, unsigned);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(int, int, unsigned, unsigned, int64, int);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(float, int, unsigned, float, double, float);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(uint64, int64, uint64, uint64, void, uint64);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(int64, int64, uint64, uint64, void, int64);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(double, int64, uint64, double, void, double);
+
+#ifndef CV_DOXYGEN
+
+#ifndef CV_CPU_OPTIMIZATION_HAL_NAMESPACE
+#ifdef CV_FORCE_SIMD128_CPP
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE hal_EMULATOR_CPP
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN namespace hal_EMULATOR_CPP {
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END }
+#elif defined(CV_CPU_DISPATCH_MODE)
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE __CV_CAT(hal_, CV_CPU_DISPATCH_MODE)
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN namespace __CV_CAT(hal_, CV_CPU_DISPATCH_MODE) {
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END }
+#else
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE hal_baseline
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN namespace hal_baseline {
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END }
+#endif
+#endif // CV_CPU_OPTIMIZATION_HAL_NAMESPACE
+
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
+using namespace CV_CPU_OPTIMIZATION_HAL_NAMESPACE;
+#endif
+}
+
+#ifdef CV_DOXYGEN
+#   undef CV_AVX2
+#   undef CV_SSE2
+#   undef CV_NEON
+#   undef CV_VSX
+#   undef CV_FP16
+#   undef CV_MSA
+#endif
+
+#if (CV_SSE2 || CV_NEON || CV_VSX || CV_MSA || CV_WASM_SIMD) && !defined(CV_FORCE_SIMD128_CPP)
+#define CV__SIMD_FORWARD 128
+#include "opencv2/core/hal/intrin_forward.hpp"
+#endif
+
+#if CV_SSE2 && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_sse_em.hpp"
+#include "opencv2/core/hal/intrin_sse.hpp"
+
+#elif CV_NEON && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_neon.hpp"
+
+#elif CV_VSX && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_vsx.hpp"
+
+#elif CV_MSA && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_msa.hpp"
+
+#elif CV_WASM_SIMD && !defined(CV_FORCE_SIMD128_CPP)
+#include "opencv2/core/hal/intrin_wasm.hpp"
+
+#else
+
+#include "opencv2/core/hal/intrin_cpp.hpp"
+
+#endif
+
+// AVX2 can be used together with SSE2, so
+// we define those two sets of intrinsics at once.
+// Most of the intrinsics do not conflict (the proper overloaded variant is
+// resolved by the argument types, e.g. v_float32x4 ~ SSE2, v_float32x8 ~ AVX2),
+// but some of AVX2 intrinsics get v256_ prefix instead of v_, e.g. v256_load() vs v_load().
+// Correspondingly, the wide intrinsics (which are mapped to the "widest"
+// available instruction set) will get vx_ prefix
+// (and will be mapped to v256_ counterparts) (e.g. vx_load() => v256_load())
+#if CV_AVX2
+
+#define CV__SIMD_FORWARD 256
+#include "opencv2/core/hal/intrin_forward.hpp"
+#include "opencv2/core/hal/intrin_avx.hpp"
+
+#endif
+
+// AVX512 can be used together with SSE2 and AVX2, so
+// we define those sets of intrinsics at once.
+// For some of AVX512 intrinsics get v512_ prefix instead of v_, e.g. v512_load() vs v_load().
+// Wide intrinsics will be mapped to v512_ counterparts in this case(e.g. vx_load() => v512_load())
+#if CV_AVX512_SKX
+
+#define CV__SIMD_FORWARD 512
+#include "opencv2/core/hal/intrin_forward.hpp"
+#include "opencv2/core/hal/intrin_avx512.hpp"
+
+#endif
+
+//! @cond IGNORED
+
+namespace cv {
+
+#ifndef CV_DOXYGEN
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN
+#endif
+
+#ifndef CV_SIMD128
+#define CV_SIMD128 0
+#endif
+
+#ifndef CV_SIMD128_CPP
+#define CV_SIMD128_CPP 0
+#endif
+
+#ifndef CV_SIMD128_64F
+#define CV_SIMD128_64F 0
+#endif
+
+#ifndef CV_SIMD256
+#define CV_SIMD256 0
+#endif
+
+#ifndef CV_SIMD256_64F
+#define CV_SIMD256_64F 0
+#endif
+
+#ifndef CV_SIMD512
+#define CV_SIMD512 0
+#endif
+
+#ifndef CV_SIMD512_64F
+#define CV_SIMD512_64F 0
+#endif
+
+#ifndef CV_SIMD128_FP16
+#define CV_SIMD128_FP16 0
+#endif
+
+#ifndef CV_SIMD256_FP16
+#define CV_SIMD256_FP16 0
+#endif
+
+#ifndef CV_SIMD512_FP16
+#define CV_SIMD512_FP16 0
+#endif
+
+//==================================================================================================
+
+template<typename _Tp> struct V_RegTraits
+{
+};
+
+#define CV_DEF_REG_TRAITS(prefix, _reg, lane_type, suffix, _u_reg, _w_reg, _q_reg, _int_reg, _round_reg) \
+    template<> struct V_RegTraits<_reg> \
+    { \
+        typedef _reg reg; \
+        typedef _u_reg u_reg; \
+        typedef _w_reg w_reg; \
+        typedef _q_reg q_reg; \
+        typedef _int_reg int_reg; \
+        typedef _round_reg round_reg; \
+    }
+
+#if CV_SIMD128 || CV_SIMD128_CPP
+    CV_DEF_REG_TRAITS(v, v_uint8x16, uchar, u8, v_uint8x16, v_uint16x8, v_uint32x4, v_int8x16, void);
+    CV_DEF_REG_TRAITS(v, v_int8x16, schar, s8, v_uint8x16, v_int16x8, v_int32x4, v_int8x16, void);
+    CV_DEF_REG_TRAITS(v, v_uint16x8, ushort, u16, v_uint16x8, v_uint32x4, v_uint64x2, v_int16x8, void);
+    CV_DEF_REG_TRAITS(v, v_int16x8, short, s16, v_uint16x8, v_int32x4, v_int64x2, v_int16x8, void);
+    CV_DEF_REG_TRAITS(v, v_uint32x4, unsigned, u32, v_uint32x4, v_uint64x2, void, v_int32x4, void);
+    CV_DEF_REG_TRAITS(v, v_int32x4, int, s32, v_uint32x4, v_int64x2, void, v_int32x4, void);
+#if CV_SIMD128_64F || CV_SIMD128_CPP
+    CV_DEF_REG_TRAITS(v, v_float32x4, float, f32, v_float32x4, v_float64x2, void, v_int32x4, v_int32x4);
+#else
+    CV_DEF_REG_TRAITS(v, v_float32x4, float, f32, v_float32x4, void, void, v_int32x4, v_int32x4);
+#endif
+    CV_DEF_REG_TRAITS(v, v_uint64x2, uint64, u64, v_uint64x2, void, void, v_int64x2, void);
+    CV_DEF_REG_TRAITS(v, v_int64x2, int64, s64, v_uint64x2, void, void, v_int64x2, void);
+#if CV_SIMD128_64F
+    CV_DEF_REG_TRAITS(v, v_float64x2, double, f64, v_float64x2, void, void, v_int64x2, v_int32x4);
+#endif
+#endif
+
+#if CV_SIMD256
+    CV_DEF_REG_TRAITS(v256, v_uint8x32, uchar, u8, v_uint8x32, v_uint16x16, v_uint32x8, v_int8x32, void);
+    CV_DEF_REG_TRAITS(v256, v_int8x32, schar, s8, v_uint8x32, v_int16x16, v_int32x8, v_int8x32, void);
+    CV_DEF_REG_TRAITS(v256, v_uint16x16, ushort, u16, v_uint16x16, v_uint32x8, v_uint64x4, v_int16x16, void);
+    CV_DEF_REG_TRAITS(v256, v_int16x16, short, s16, v_uint16x16, v_int32x8, v_int64x4, v_int16x16, void);
+    CV_DEF_REG_TRAITS(v256, v_uint32x8, unsigned, u32, v_uint32x8, v_uint64x4, void, v_int32x8, void);
+    CV_DEF_REG_TRAITS(v256, v_int32x8, int, s32, v_uint32x8, v_int64x4, void, v_int32x8, void);
+    CV_DEF_REG_TRAITS(v256, v_float32x8, float, f32, v_float32x8, v_float64x4, void, v_int32x8, v_int32x8);
+    CV_DEF_REG_TRAITS(v256, v_uint64x4, uint64, u64, v_uint64x4, void, void, v_int64x4, void);
+    CV_DEF_REG_TRAITS(v256, v_int64x4, int64, s64, v_uint64x4, void, void, v_int64x4, void);
+    CV_DEF_REG_TRAITS(v256, v_float64x4, double, f64, v_float64x4, void, void, v_int64x4, v_int32x8);
+#endif
+
+#if CV_SIMD512
+    CV_DEF_REG_TRAITS(v512, v_uint8x64, uchar, u8, v_uint8x64, v_uint16x32, v_uint32x16, v_int8x64, void);
+    CV_DEF_REG_TRAITS(v512, v_int8x64, schar, s8, v_uint8x64, v_int16x32, v_int32x16, v_int8x64, void);
+    CV_DEF_REG_TRAITS(v512, v_uint16x32, ushort, u16, v_uint16x32, v_uint32x16, v_uint64x8, v_int16x32, void);
+    CV_DEF_REG_TRAITS(v512, v_int16x32, short, s16, v_uint16x32, v_int32x16, v_int64x8, v_int16x32, void);
+    CV_DEF_REG_TRAITS(v512, v_uint32x16, unsigned, u32, v_uint32x16, v_uint64x8, void, v_int32x16, void);
+    CV_DEF_REG_TRAITS(v512, v_int32x16, int, s32, v_uint32x16, v_int64x8, void, v_int32x16, void);
+    CV_DEF_REG_TRAITS(v512, v_float32x16, float, f32, v_float32x16, v_float64x8, void, v_int32x16, v_int32x16);
+    CV_DEF_REG_TRAITS(v512, v_uint64x8, uint64, u64, v_uint64x8, void, void, v_int64x8, void);
+    CV_DEF_REG_TRAITS(v512, v_int64x8, int64, s64, v_uint64x8, void, void, v_int64x8, void);
+    CV_DEF_REG_TRAITS(v512, v_float64x8, double, f64, v_float64x8, void, void, v_int64x8, v_int32x16);
+#endif
+//! @endcond
+
+#if CV_SIMD512 && (!defined(CV__SIMD_FORCE_WIDTH) || CV__SIMD_FORCE_WIDTH == 512)
+#define CV__SIMD_NAMESPACE simd512
+namespace CV__SIMD_NAMESPACE {
+    #define CV_SIMD 1
+    #define CV_SIMD_64F CV_SIMD512_64F
+    #define CV_SIMD_FP16 CV_SIMD512_FP16
+    #define CV_SIMD_WIDTH 64
+//! @addtogroup core_hal_intrin
+//! @{
+    //! @brief Maximum available vector register capacity 8-bit unsigned integer values
+    typedef v_uint8x64    v_uint8;
+    //! @brief Maximum available vector register capacity 8-bit signed integer values
+    typedef v_int8x64     v_int8;
+    //! @brief Maximum available vector register capacity 16-bit unsigned integer values
+    typedef v_uint16x32   v_uint16;
+    //! @brief Maximum available vector register capacity 16-bit signed integer values
+    typedef v_int16x32    v_int16;
+    //! @brief Maximum available vector register capacity 32-bit unsigned integer values
+    typedef v_uint32x16   v_uint32;
+    //! @brief Maximum available vector register capacity 32-bit signed integer values
+    typedef v_int32x16    v_int32;
+    //! @brief Maximum available vector register capacity 64-bit unsigned integer values
+    typedef v_uint64x8    v_uint64;
+    //! @brief Maximum available vector register capacity 64-bit signed integer values
+    typedef v_int64x8     v_int64;
+    //! @brief Maximum available vector register capacity 32-bit floating point values (single precision)
+    typedef v_float32x16  v_float32;
+    #if CV_SIMD512_64F
+    //! @brief Maximum available vector register capacity 64-bit floating point values (double precision)
+    typedef v_float64x8   v_float64;
+    #endif
+//! @}
+
+    #define VXPREFIX(func) v512##func
+} // namespace
+using namespace CV__SIMD_NAMESPACE;
+#elif CV_SIMD256 && (!defined(CV__SIMD_FORCE_WIDTH) || CV__SIMD_FORCE_WIDTH == 256)
+#define CV__SIMD_NAMESPACE simd256
+namespace CV__SIMD_NAMESPACE {
+    #define CV_SIMD 1
+    #define CV_SIMD_64F CV_SIMD256_64F
+    #define CV_SIMD_FP16 CV_SIMD256_FP16
+    #define CV_SIMD_WIDTH 32
+//! @addtogroup core_hal_intrin
+//! @{
+    //! @brief Maximum available vector register capacity 8-bit unsigned integer values
+    typedef v_uint8x32   v_uint8;
+    //! @brief Maximum available vector register capacity 8-bit signed integer values
+    typedef v_int8x32    v_int8;
+    //! @brief Maximum available vector register capacity 16-bit unsigned integer values
+    typedef v_uint16x16  v_uint16;
+    //! @brief Maximum available vector register capacity 16-bit signed integer values
+    typedef v_int16x16   v_int16;
+    //! @brief Maximum available vector register capacity 32-bit unsigned integer values
+    typedef v_uint32x8   v_uint32;
+    //! @brief Maximum available vector register capacity 32-bit signed integer values
+    typedef v_int32x8    v_int32;
+    //! @brief Maximum available vector register capacity 64-bit unsigned integer values
+    typedef v_uint64x4   v_uint64;
+    //! @brief Maximum available vector register capacity 64-bit signed integer values
+    typedef v_int64x4    v_int64;
+    //! @brief Maximum available vector register capacity 32-bit floating point values (single precision)
+    typedef v_float32x8  v_float32;
+    #if CV_SIMD256_64F
+    //! @brief Maximum available vector register capacity 64-bit floating point values (double precision)
+    typedef v_float64x4  v_float64;
+    #endif
+//! @}
+
+    #define VXPREFIX(func) v256##func
+} // namespace
+using namespace CV__SIMD_NAMESPACE;
+#elif (CV_SIMD128 || CV_SIMD128_CPP) && (!defined(CV__SIMD_FORCE_WIDTH) || CV__SIMD_FORCE_WIDTH == 128)
+#if defined CV_SIMD128_CPP
+#define CV__SIMD_NAMESPACE simd128_cpp
+#else
+#define CV__SIMD_NAMESPACE simd128
+#endif
+namespace CV__SIMD_NAMESPACE {
+    #define CV_SIMD CV_SIMD128
+    #define CV_SIMD_64F CV_SIMD128_64F
+    #define CV_SIMD_WIDTH 16
+//! @addtogroup core_hal_intrin
+//! @{
+    //! @brief Maximum available vector register capacity 8-bit unsigned integer values
+    typedef v_uint8x16  v_uint8;
+    //! @brief Maximum available vector register capacity 8-bit signed integer values
+    typedef v_int8x16   v_int8;
+    //! @brief Maximum available vector register capacity 16-bit unsigned integer values
+    typedef v_uint16x8  v_uint16;
+    //! @brief Maximum available vector register capacity 16-bit signed integer values
+    typedef v_int16x8   v_int16;
+    //! @brief Maximum available vector register capacity 32-bit unsigned integer values
+    typedef v_uint32x4  v_uint32;
+    //! @brief Maximum available vector register capacity 32-bit signed integer values
+    typedef v_int32x4   v_int32;
+    //! @brief Maximum available vector register capacity 64-bit unsigned integer values
+    typedef v_uint64x2  v_uint64;
+    //! @brief Maximum available vector register capacity 64-bit signed integer values
+    typedef v_int64x2   v_int64;
+    //! @brief Maximum available vector register capacity 32-bit floating point values (single precision)
+    typedef v_float32x4 v_float32;
+    #if CV_SIMD128_64F
+    //! @brief Maximum available vector register capacity 64-bit floating point values (double precision)
+    typedef v_float64x2 v_float64;
+    #endif
+//! @}
+
+    #define VXPREFIX(func) v##func
+} // namespace
+using namespace CV__SIMD_NAMESPACE;
+#endif
+
+namespace CV__SIMD_NAMESPACE {
+//! @addtogroup core_hal_intrin
+//! @{
+    //! @name Wide init with value
+    //! @{
+    //! @brief Create maximum available capacity vector with elements set to a specific value
+    inline v_uint8 vx_setall_u8(uchar v) { return VXPREFIX(_setall_u8)(v); }
+    inline v_int8 vx_setall_s8(schar v) { return VXPREFIX(_setall_s8)(v); }
+    inline v_uint16 vx_setall_u16(ushort v) { return VXPREFIX(_setall_u16)(v); }
+    inline v_int16 vx_setall_s16(short v) { return VXPREFIX(_setall_s16)(v); }
+    inline v_int32 vx_setall_s32(int v) { return VXPREFIX(_setall_s32)(v); }
+    inline v_uint32 vx_setall_u32(unsigned v) { return VXPREFIX(_setall_u32)(v); }
+    inline v_float32 vx_setall_f32(float v) { return VXPREFIX(_setall_f32)(v); }
+    inline v_int64 vx_setall_s64(int64 v) { return VXPREFIX(_setall_s64)(v); }
+    inline v_uint64 vx_setall_u64(uint64 v) { return VXPREFIX(_setall_u64)(v); }
+#if CV_SIMD_64F
+    inline v_float64 vx_setall_f64(double v) { return VXPREFIX(_setall_f64)(v); }
+#endif
+    //! @}
+
+    //! @name Wide init with zero
+    //! @{
+    //! @brief Create maximum available capacity vector with elements set to zero
+    inline v_uint8 vx_setzero_u8() { return VXPREFIX(_setzero_u8)(); }
+    inline v_int8 vx_setzero_s8() { return VXPREFIX(_setzero_s8)(); }
+    inline v_uint16 vx_setzero_u16() { return VXPREFIX(_setzero_u16)(); }
+    inline v_int16 vx_setzero_s16() { return VXPREFIX(_setzero_s16)(); }
+    inline v_int32 vx_setzero_s32() { return VXPREFIX(_setzero_s32)(); }
+    inline v_uint32 vx_setzero_u32() { return VXPREFIX(_setzero_u32)(); }
+    inline v_float32 vx_setzero_f32() { return VXPREFIX(_setzero_f32)(); }
+    inline v_int64 vx_setzero_s64() { return VXPREFIX(_setzero_s64)(); }
+    inline v_uint64 vx_setzero_u64() { return VXPREFIX(_setzero_u64)(); }
+#if CV_SIMD_64F
+    inline v_float64 vx_setzero_f64() { return VXPREFIX(_setzero_f64)(); }
+#endif
+    //! @}
+
+    //! @name Wide load from memory
+    //! @{
+    //! @brief Load maximum available capacity register contents from memory
+    inline v_uint8 vx_load(const uchar * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_int8 vx_load(const schar * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_uint16 vx_load(const ushort * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_int16 vx_load(const short * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_int32 vx_load(const int * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_uint32 vx_load(const unsigned * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_float32 vx_load(const float * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_int64 vx_load(const int64 * ptr) { return VXPREFIX(_load)(ptr); }
+    inline v_uint64 vx_load(const uint64 * ptr) { return VXPREFIX(_load)(ptr); }
+#if CV_SIMD_64F
+    inline v_float64 vx_load(const double * ptr) { return VXPREFIX(_load)(ptr); }
+#endif
+    //! @}
+
+    //! @name Wide load from memory(aligned)
+    //! @{
+    //! @brief Load maximum available capacity register contents from memory(aligned)
+    inline v_uint8 vx_load_aligned(const uchar * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_int8 vx_load_aligned(const schar * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_uint16 vx_load_aligned(const ushort * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_int16 vx_load_aligned(const short * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_int32 vx_load_aligned(const int * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_uint32 vx_load_aligned(const unsigned * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_float32 vx_load_aligned(const float * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_int64 vx_load_aligned(const int64 * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+    inline v_uint64 vx_load_aligned(const uint64 * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+#if CV_SIMD_64F
+    inline v_float64 vx_load_aligned(const double * ptr) { return VXPREFIX(_load_aligned)(ptr); }
+#endif
+    //! @}
+
+    //! @name Wide load lower half from memory
+    //! @{
+    //! @brief Load lower half of maximum available capacity register from memory
+    inline v_uint8 vx_load_low(const uchar * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_int8 vx_load_low(const schar * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_uint16 vx_load_low(const ushort * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_int16 vx_load_low(const short * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_int32 vx_load_low(const int * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_uint32 vx_load_low(const unsigned * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_float32 vx_load_low(const float * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_int64 vx_load_low(const int64 * ptr) { return VXPREFIX(_load_low)(ptr); }
+    inline v_uint64 vx_load_low(const uint64 * ptr) { return VXPREFIX(_load_low)(ptr); }
+#if CV_SIMD_64F
+    inline v_float64 vx_load_low(const double * ptr) { return VXPREFIX(_load_low)(ptr); }
+#endif
+    //! @}
+
+    //! @name Wide load halfs from memory
+    //! @{
+    //! @brief Load maximum available capacity register contents from two memory blocks
+    inline v_uint8 vx_load_halves(const uchar * ptr0, const uchar * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_int8 vx_load_halves(const schar * ptr0, const schar * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_uint16 vx_load_halves(const ushort * ptr0, const ushort * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_int16 vx_load_halves(const short * ptr0, const short * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_int32 vx_load_halves(const int * ptr0, const int * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_uint32 vx_load_halves(const unsigned * ptr0, const unsigned * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_float32 vx_load_halves(const float * ptr0, const float * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_int64 vx_load_halves(const int64 * ptr0, const int64 * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+    inline v_uint64 vx_load_halves(const uint64 * ptr0, const uint64 * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+#if CV_SIMD_64F
+    inline v_float64 vx_load_halves(const double * ptr0, const double * ptr1) { return VXPREFIX(_load_halves)(ptr0, ptr1); }
+#endif
+    //! @}
+
+    //! @name Wide LUT of elements
+    //! @{
+    //! @brief Load maximum available capacity register contents with array elements by provided indexes
+    inline v_uint8 vx_lut(const uchar * ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_int8 vx_lut(const schar * ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_uint16 vx_lut(const ushort * ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_int16 vx_lut(const short* ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_int32 vx_lut(const int* ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_uint32 vx_lut(const unsigned* ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_float32 vx_lut(const float* ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_int64 vx_lut(const int64 * ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+    inline v_uint64 vx_lut(const uint64 * ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+#if CV_SIMD_64F
+    inline v_float64 vx_lut(const double* ptr, const int* idx) { return VXPREFIX(_lut)(ptr, idx); }
+#endif
+    //! @}
+
+    //! @name Wide LUT of element pairs
+    //! @{
+    //! @brief Load maximum available capacity register contents with array element pairs by provided indexes
+    inline v_uint8 vx_lut_pairs(const uchar * ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_int8 vx_lut_pairs(const schar * ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_uint16 vx_lut_pairs(const ushort * ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_int16 vx_lut_pairs(const short* ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_int32 vx_lut_pairs(const int* ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_uint32 vx_lut_pairs(const unsigned* ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_float32 vx_lut_pairs(const float* ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_int64 vx_lut_pairs(const int64 * ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+    inline v_uint64 vx_lut_pairs(const uint64 * ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+#if CV_SIMD_64F
+    inline v_float64 vx_lut_pairs(const double* ptr, const int* idx) { return VXPREFIX(_lut_pairs)(ptr, idx); }
+#endif
+    //! @}
+
+    //! @name Wide LUT of element quads
+    //! @{
+    //! @brief Load maximum available capacity register contents with array element quads by provided indexes
+    inline v_uint8 vx_lut_quads(const uchar* ptr, const int* idx) { return VXPREFIX(_lut_quads)(ptr, idx); }
+    inline v_int8 vx_lut_quads(const schar* ptr, const int* idx) { return VXPREFIX(_lut_quads)(ptr, idx); }
+    inline v_uint16 vx_lut_quads(const ushort* ptr, const int* idx) { return VXPREFIX(_lut_quads)(ptr, idx); }
+    inline v_int16 vx_lut_quads(const short* ptr, const int* idx) { return VXPREFIX(_lut_quads)(ptr, idx); }
+    inline v_int32 vx_lut_quads(const int* ptr, const int* idx) { return VXPREFIX(_lut_quads)(ptr, idx); }
+    inline v_uint32 vx_lut_quads(const unsigned* ptr, const int* idx) { return VXPREFIX(_lut_quads)(ptr, idx); }
+    inline v_float32 vx_lut_quads(const float* ptr, const int* idx) { return VXPREFIX(_lut_quads)(ptr, idx); }
+    //! @}
+
+    //! @name Wide load with double expansion
+    //! @{
+    //! @brief Load maximum available capacity register contents from memory with double expand
+    inline v_uint16 vx_load_expand(const uchar * ptr) { return VXPREFIX(_load_expand)(ptr); }
+    inline v_int16 vx_load_expand(const schar * ptr) { return VXPREFIX(_load_expand)(ptr); }
+    inline v_uint32 vx_load_expand(const ushort * ptr) { return VXPREFIX(_load_expand)(ptr); }
+    inline v_int32 vx_load_expand(const short* ptr) { return VXPREFIX(_load_expand)(ptr); }
+    inline v_int64 vx_load_expand(const int* ptr) { return VXPREFIX(_load_expand)(ptr); }
+    inline v_uint64 vx_load_expand(const unsigned* ptr) { return VXPREFIX(_load_expand)(ptr); }
+    inline v_float32 vx_load_expand(const float16_t * ptr) { return VXPREFIX(_load_expand)(ptr); }
+    //! @}
+
+    //! @name Wide load with quad expansion
+    //! @{
+    //! @brief Load maximum available capacity register contents from memory with quad expand
+    inline v_uint32 vx_load_expand_q(const uchar * ptr) { return VXPREFIX(_load_expand_q)(ptr); }
+    inline v_int32 vx_load_expand_q(const schar * ptr) { return VXPREFIX(_load_expand_q)(ptr); }
+    //! @}
+
+    /** @brief SIMD processing state cleanup call */
+    inline void vx_cleanup() { VXPREFIX(_cleanup)(); }
+
+
+//! @cond IGNORED
+
+    // backward compatibility
+    template<typename _Tp, typename _Tvec> static inline
+    void vx_store(_Tp* dst, const _Tvec& v) { return v_store(dst, v); }
+    // backward compatibility
+    template<typename _Tp, typename _Tvec> static inline
+    void vx_store_aligned(_Tp* dst, const _Tvec& v) { return v_store_aligned(dst, v); }
+
+//! @endcond
+
+
+//! @}
+    #undef VXPREFIX
+} // namespace
+
+//! @cond IGNORED
+#ifndef CV_SIMD_64F
+#define CV_SIMD_64F 0
+#endif
+
+#ifndef CV_SIMD_FP16
+#define CV_SIMD_FP16 0  //!< Defined to 1 on native support of operations with float16x8_t / float16x16_t (SIMD256) types
+#endif
+
+#ifndef CV_SIMD
+#define CV_SIMD 0
+#endif
+
+#include "simd_utils.impl.hpp"
+
+#ifndef CV_DOXYGEN
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
+#endif
+
+} // cv::
+
+//! @endcond
+
+#endif

3rdPart/include/opencv2/core/hal/intrin_forward.hpp

@@ -0,0 +1,191 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html
+
+#ifndef CV__SIMD_FORWARD
+#error "Need to pre-define forward width"
+#endif
+
+namespace cv
+{
+
+//! @cond IGNORED
+
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN
+
+/** Types **/
+#if CV__SIMD_FORWARD == 1024
+// [todo] 1024
+#error "1024-long ops not implemented yet"
+#elif CV__SIMD_FORWARD == 512
+// 512
+#define __CV_VX(fun)   v512_##fun
+#define __CV_V_UINT8   v_uint8x64
+#define __CV_V_INT8    v_int8x64
+#define __CV_V_UINT16  v_uint16x32
+#define __CV_V_INT16   v_int16x32
+#define __CV_V_UINT32  v_uint32x16
+#define __CV_V_INT32   v_int32x16
+#define __CV_V_UINT64  v_uint64x8
+#define __CV_V_INT64   v_int64x8
+#define __CV_V_FLOAT32 v_float32x16
+#define __CV_V_FLOAT64 v_float64x8
+struct v_uint8x64;
+struct v_int8x64;
+struct v_uint16x32;
+struct v_int16x32;
+struct v_uint32x16;
+struct v_int32x16;
+struct v_uint64x8;
+struct v_int64x8;
+struct v_float32x16;
+struct v_float64x8;
+#elif CV__SIMD_FORWARD == 256
+// 256
+#define __CV_VX(fun)   v256_##fun
+#define __CV_V_UINT8   v_uint8x32
+#define __CV_V_INT8    v_int8x32
+#define __CV_V_UINT16  v_uint16x16
+#define __CV_V_INT16   v_int16x16
+#define __CV_V_UINT32  v_uint32x8
+#define __CV_V_INT32   v_int32x8
+#define __CV_V_UINT64  v_uint64x4
+#define __CV_V_INT64   v_int64x4
+#define __CV_V_FLOAT32 v_float32x8
+#define __CV_V_FLOAT64 v_float64x4
+struct v_uint8x32;
+struct v_int8x32;
+struct v_uint16x16;
+struct v_int16x16;
+struct v_uint32x8;
+struct v_int32x8;
+struct v_uint64x4;
+struct v_int64x4;
+struct v_float32x8;
+struct v_float64x4;
+#else
+// 128
+#define __CV_VX(fun)   v_##fun
+#define __CV_V_UINT8   v_uint8x16
+#define __CV_V_INT8    v_int8x16
+#define __CV_V_UINT16  v_uint16x8
+#define __CV_V_INT16   v_int16x8
+#define __CV_V_UINT32  v_uint32x4
+#define __CV_V_INT32   v_int32x4
+#define __CV_V_UINT64  v_uint64x2
+#define __CV_V_INT64   v_int64x2
+#define __CV_V_FLOAT32 v_float32x4
+#define __CV_V_FLOAT64 v_float64x2
+struct v_uint8x16;
+struct v_int8x16;
+struct v_uint16x8;
+struct v_int16x8;
+struct v_uint32x4;
+struct v_int32x4;
+struct v_uint64x2;
+struct v_int64x2;
+struct v_float32x4;
+struct v_float64x2;
+#endif
+
+/** Value reordering **/
+
+// Expansion
+void v_expand(const __CV_V_UINT8&,  __CV_V_UINT16&, __CV_V_UINT16&);
+void v_expand(const __CV_V_INT8&,   __CV_V_INT16&,  __CV_V_INT16&);
+void v_expand(const __CV_V_UINT16&, __CV_V_UINT32&, __CV_V_UINT32&);
+void v_expand(const __CV_V_INT16&,  __CV_V_INT32&,  __CV_V_INT32&);
+void v_expand(const __CV_V_UINT32&, __CV_V_UINT64&, __CV_V_UINT64&);
+void v_expand(const __CV_V_INT32&,  __CV_V_INT64&,  __CV_V_INT64&);
+// Low Expansion
+__CV_V_UINT16 v_expand_low(const __CV_V_UINT8&);
+__CV_V_INT16  v_expand_low(const __CV_V_INT8&);
+__CV_V_UINT32 v_expand_low(const __CV_V_UINT16&);
+__CV_V_INT32  v_expand_low(const __CV_V_INT16&);
+__CV_V_UINT64 v_expand_low(const __CV_V_UINT32&);
+__CV_V_INT64  v_expand_low(const __CV_V_INT32&);
+// High Expansion
+__CV_V_UINT16 v_expand_high(const __CV_V_UINT8&);
+__CV_V_INT16  v_expand_high(const __CV_V_INT8&);
+__CV_V_UINT32 v_expand_high(const __CV_V_UINT16&);
+__CV_V_INT32  v_expand_high(const __CV_V_INT16&);
+__CV_V_UINT64 v_expand_high(const __CV_V_UINT32&);
+__CV_V_INT64  v_expand_high(const __CV_V_INT32&);
+// Load & Low Expansion
+__CV_V_UINT16 __CV_VX(load_expand)(const uchar*);
+__CV_V_INT16  __CV_VX(load_expand)(const schar*);
+__CV_V_UINT32 __CV_VX(load_expand)(const ushort*);
+__CV_V_INT32  __CV_VX(load_expand)(const short*);
+__CV_V_UINT64 __CV_VX(load_expand)(const uint*);
+__CV_V_INT64  __CV_VX(load_expand)(const int*);
+// Load lower 8-bit and expand into 32-bit
+__CV_V_UINT32 __CV_VX(load_expand_q)(const uchar*);
+__CV_V_INT32  __CV_VX(load_expand_q)(const schar*);
+
+// Saturating Pack
+__CV_V_UINT8  v_pack(const __CV_V_UINT16&, const __CV_V_UINT16&);
+__CV_V_INT8   v_pack(const __CV_V_INT16&,  const __CV_V_INT16&);
+__CV_V_UINT16 v_pack(const __CV_V_UINT32&, const __CV_V_UINT32&);
+__CV_V_INT16  v_pack(const __CV_V_INT32&,  const __CV_V_INT32&);
+// Non-saturating Pack
+__CV_V_UINT32 v_pack(const __CV_V_UINT64&, const __CV_V_UINT64&);
+__CV_V_INT32  v_pack(const __CV_V_INT64&,  const __CV_V_INT64&);
+// Pack signed integers with unsigned saturation
+__CV_V_UINT8  v_pack_u(const __CV_V_INT16&, const __CV_V_INT16&);
+__CV_V_UINT16 v_pack_u(const __CV_V_INT32&, const __CV_V_INT32&);
+
+/** Arithmetic, bitwise and comparison operations **/
+
+// Non-saturating multiply
+#if CV_VSX
+template<typename Tvec>
+Tvec v_mul_wrap(const Tvec& a, const Tvec& b);
+#else
+__CV_V_UINT8  v_mul_wrap(const __CV_V_UINT8&,  const __CV_V_UINT8&);
+__CV_V_INT8   v_mul_wrap(const __CV_V_INT8&,   const __CV_V_INT8&);
+__CV_V_UINT16 v_mul_wrap(const __CV_V_UINT16&, const __CV_V_UINT16&);
+__CV_V_INT16  v_mul_wrap(const __CV_V_INT16&,  const __CV_V_INT16&);
+#endif
+
+//  Multiply and expand
+#if CV_VSX
+template<typename Tvec, typename Twvec>
+void v_mul_expand(const Tvec& a, const Tvec& b, Twvec& c, Twvec& d);
+#else
+void v_mul_expand(const __CV_V_UINT8&,  const __CV_V_UINT8&,  __CV_V_UINT16&, __CV_V_UINT16&);
+void v_mul_expand(const __CV_V_INT8&,   const __CV_V_INT8&,   __CV_V_INT16&,  __CV_V_INT16&);
+void v_mul_expand(const __CV_V_UINT16&, const __CV_V_UINT16&, __CV_V_UINT32&, __CV_V_UINT32&);
+void v_mul_expand(const __CV_V_INT16&,  const __CV_V_INT16&,  __CV_V_INT32&,  __CV_V_INT32&);
+void v_mul_expand(const __CV_V_UINT32&, const __CV_V_UINT32&, __CV_V_UINT64&, __CV_V_UINT64&);
+void v_mul_expand(const __CV_V_INT32&,  const __CV_V_INT32&,  __CV_V_INT64&,  __CV_V_INT64&);
+#endif
+
+// Conversions
+__CV_V_FLOAT32 v_cvt_f32(const __CV_V_INT32& a);
+__CV_V_FLOAT32 v_cvt_f32(const __CV_V_FLOAT64& a);
+__CV_V_FLOAT32 v_cvt_f32(const __CV_V_FLOAT64& a, const __CV_V_FLOAT64& b);
+__CV_V_FLOAT64 v_cvt_f64(const __CV_V_INT32& a);
+__CV_V_FLOAT64 v_cvt_f64_high(const __CV_V_INT32& a);
+__CV_V_FLOAT64 v_cvt_f64(const __CV_V_FLOAT32& a);
+__CV_V_FLOAT64 v_cvt_f64_high(const __CV_V_FLOAT32& a);
+__CV_V_FLOAT64 v_cvt_f64(const __CV_V_INT64& a);
+
+/** Cleanup **/
+#undef CV__SIMD_FORWARD
+#undef __CV_VX
+#undef __CV_V_UINT8
+#undef __CV_V_INT8
+#undef __CV_V_UINT16
+#undef __CV_V_INT16
+#undef __CV_V_UINT32
+#undef __CV_V_INT32
+#undef __CV_V_UINT64
+#undef __CV_V_INT64
+#undef __CV_V_FLOAT32
+#undef __CV_V_FLOAT64
+
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
+
+//! @endcond
+
+} // cv::