#include "stdafx.h"

#include "cv.h"

#include "highgui.h"

#include <stdio.h>

#include <stdlib.h>

#include <omp.h>

#pragma comment(lib,"opencv_core2410d.lib")

#pragma comment(lib,"opencv_highgui2410d.lib")

#pragma comment(lib,"opencv_imgproc2410d.lib")    

void EdgeOpenMP(IplImage *src,IplImage *dst,int thresh)

{

    int height    = src->height;

    int width     = src->width;

    int step      = src->widthStep;

    uchar *data1      = (uchar *)src->imageData;

    uchar *data2      = (uchar *)dst->imageData;

    int i=step;

    #pragma omp parallel for

    for(i=step+1;i<height*width;i++){

         if(abs(data1[i]-data1[i-1])>thresh || abs(data1[i]-data1[i-step])>thresh)

            data2[i]=255;/* 对于单通道，前后两帧差分大于门限

            或者对于多通道前后两帧的一个指标差分大于门限，则视为边缘*/

         else

            data2[i]=0;

    }

}

void Edge(IplImage *src,IplImage *dst,int thresh)

{

    int height    = src->height;

    int width     = src->width;

    int step      = src->widthStep;

    uchar *data1      = (uchar *)src->imageData;

    uchar *data2      = (uchar *)dst->imageData;

   int i=step;

    for(i=step+1;i<height*width;i++){

         if(abs(data1[i]-data1[i-1])>thresh || abs(data1[i]-data1[i-step])>thresh)

            data2[i]=255;

         else

            data2[i]=0;

    }

}

int main()

{

  char filename[512];

  IplImage *src,*edge1,*edge2;

  puts("File name:");

  gets(filename);

  src = cvLoadImage(filename,CV_LOAD_IMAGE_GRAYSCALE );

  edge1=cvCloneImage(src);

  edge2=cvCloneImage(src);

  cvNamedWindow("src", CV_WINDOW_AUTOSIZE);

  cvMoveWindow("src", 100, 100);

  cvShowImage( "src", src);

  cvNamedWindow("Edge", CV_WINDOW_AUTOSIZE);

  cvMoveWindow("Edge", 200, 100);

  cvNamedWindow("EdgeOpenMP", CV_WINDOW_AUTOSIZE);

  cvMoveWindow("EdgeOpenMP", 300, 100);

  /* 以上都是准备一些窗口和图形基本数据 */

  int tekrar=100;//运行次数

  int thresh=30;

  double start, end,t1, t2; 

  /* 计算没有使用OpenMP优化的时间 */

  start= (double)cvGetTickCount();//记下开始的时钟计数，以便计算函数或用户代码执行时间

  for(int i=0;i<tekrar;i++)

    Edge(src,edge1,thresh);

  end= (double)cvGetTickCount();//记下结束的时钟计数

  t1= (end-start)/((double)cvGetTickFrequency()*1000.);//计算运行时间，以毫秒为单位

  printf( "Run time without OpenMP = %g ms\n", t1 );

  /* 计算使用了OpenMP优化的时间 */

  start= (double)cvGetTickCount();

  for(int i=0;i<tekrar;i++)

    EdgeOpenMP(src,edge2,thresh);

  end= (double)cvGetTickCount();

  t2= (end-start)/((double)cvGetTickFrequency()*1000.);

  printf( "Run time with OpenMP = %g ms\n", t2 );

  printf( "Performance ratio (%%) = %% %.1f \n", 100*(t1/t2-1) );

  cvShowImage( "Edge", edge1);

  cvShowImage( "EdgeOpenMP", edge2);

  cvWaitKey();

  cvDestroyWindow("Edge");

  cvDestroyWindow("EdgeOpenMP");

  cvReleaseImage(&src);

  cvReleaseImage(&edge1);

  cvReleaseImage(&edge2);

}

这是我的结果：


这里的测试结果：

http://blog.csdn.net/augusdi/article/details/8808226

  在cpp文件中添加如下代码：

#include "stdafx.h"

#include<omp.h>

#include<iostream>

usingnamespace std;

//循环测试函数

void test()

{

for(int i=0;i<10000;i++)

{

}

}

int _tmain(int argc,_TCHAR* argv[])

{

cout<<"这是一个串行测试程序!\n";

double start = omp_get_wtime( );//获取起始时间

for(int i = 0; i < 10000; i++)

{

test();

}

double end = omp_get_wtime( );

cout<<"计算耗时为："<<end -start<<"\n";

cin>>end;

return 0;

}

#include "stdafx.h"

#include<omp.h>

#include <iostream>

using namespace std;

//循环测试函数

void test()

{

for(inti=0;i<10000;i++)

{

}

}

int _tmain(int argc, _TCHAR* argv[])

{

cout<<"这是一个并行测试程序!\n";

doublestart = omp_get_wtime( );//获取起始时间

#pragma ompparallel for

for(inti = 0; i < 10000; i++)

{

test();

}

doubleend = omp_get_wtime( );

cout<<"计算耗时为："<<end -start<<"\n";

cin>>end;

return0;

}

       同样，也经过10次随机的运行，其平均耗时约为0.06358044s，两种不同运行方式的比较结果如下表所示：

次数	串行	并行
1	0.283382	0.0746704
2	0.283654	0.0686404
3	0.283212	0.0536631
4	0.280234	0.0517737
5	0.283041	0.0717588
6	0.283126	0.0524264
7	0.281881	0.0580316
8	0.283301	0.0730386
9	0.284545	0.0745088
10	0.286353	0.0572926
平均值	0.283273	0.06358044

相关程序源码下载地址： http://download.csdn.net/detail/xwebsite/3843187