#include <iostream>

#include <string.h>

#include <algorithm> //max_element and min_element

#include <math.h>

#include <stdio.h>

#include <stdlib.h>

#include <cstdio>

#include <ctime>

#include <time.h>

#include <libgen.h>

#include "opencv2/core/core.hpp"

#include "opencv2/features2d/features2d.hpp"

#include "opencv2/highgui/highgui.hpp"

#include "opencv2/calib3d/calib3d.hpp"

#include "opencv2/imgproc/imgproc.hpp"

#include "opencv2/opencv.hpp"

#include "opencv2/gpu/gpu.hpp"

#include "/usr/local/cuda/include/cuda_runtime_api.h"

using namespace std;

using namespace cv;

int main( int argc, char** argv )

{

        Mat scene = imread ( "scene.jpg",1);

        Mat templ = imread ( "template.jpg",1);

        //Split to 1 channel

        vector<Mat> rgbImg(3);

        split(scene, rgbImg);

        Mat scene1chan=rgbImg[0];

        vector<Mat> rgbImg2(3);

        split(templ, rgbImg2);

        Mat templ1chan=rgbImg2[0];

        //Upload to GPU mem

        gpu::GpuMat gpu_scene, gpu_templ;

        gpu_scene.upload(scene1chan);

        gpu_templ.upload(templ1chan);

        double minVal; double maxVal; Point minLoc; Point maxLoc;

        gpu::GpuMat gpu_result;

        gpu_result.create(gpu_scene.rows-gpu_templ.rows+1 , gpu_scene.cols-gpu_templ.cols+1 , CV_32F);

        gpu::matchTemplate(gpu_scene, gpu_templ, gpu_result,CV_TM_SQDIFF);

        gpu::minMaxLoc(gpu_result, &minVal, &maxVal, &minLoc, &maxLoc );

        Mat result=gpu_result;

        namedWindow("result", 0);

        imshow( "result", result );

        waitKey(0);

        return(0);

}