VS2010实现opencv基于DCT的图像压缩

Refer from http://blog.csdn.net/fjssharpsword/article/details/7765908

1.基于DCT的图像压缩方法是将一幅图像分割成矩形像素块，再分别对每一独立的像素块DCT变换、量化、编码和传输。

2.打开VS2010，新建Visual C++下Win32控制台应用程序demo，主程序如下：

// demo.cpp : 定义控制台应用程序的入口点。
//

#include "stdafx.h"
#include "highgui.h"
#include <math.h>
#include <cv.h>
#include "cxcore.h"
#define cvCvtPlaneToPix cvMerge 
double PSNR_B = 0;  
double PSNR_G = 0;  
double PSNR_R = 0;  
double PSNR; 

int _tmain(int argc, _TCHAR* argv[])
{
 const char* imagename = "D:/demo/demo.jpg";
 IplImage *src;  
    CvScalar SrcPixel;  
    CvScalar DstPixel;  
    double sumB = 0;  
    double sumG = 0;  
    double sumR = 0;  
    double mseB;  
    double mseG;  
    double mseR;  
 
    src= cvLoadImage( imagename,1 ) ;    
    if( !src )  
    {  
        printf("can't open the image...\n");  
        return -1;  
    }  
    // YUV颜色空间   
    IplImage* YUVImage = cvCreateImage(cvSize(src->width,src->height), src->depth, 3); 
    IplImage* dst = cvCreateImage(cvSize(src->width,src->height), src->depth, 3); 
    // YUV颜色空间各通道   
    IplImage* Y = cvCreateImage(cvSize(src->width,src->height), IPL_DEPTH_8U, 1); 
    IplImage* U = cvCreateImage(cvSize(src->width,src->height), IPL_DEPTH_8U, 1); 
    IplImage* V = cvCreateImage(cvSize(src->width,src->height), IPL_DEPTH_8U, 1); 
      
      
    //cvNamedWindow( "Origin Image", CV_WINDOW_AUTOSIZE );   
    cvCvtColor(src, YUVImage, CV_BGR2YUV); //BGR→YUV   
    cvSplit( YUVImage, Y, U, V, NULL);//分割通道   
      
    CvMat* MatY = cvCreateMat(Y->height,Y->width,CV_64FC1);  
    CvMat* MatU = cvCreateMat(V->height,U->width,CV_64FC1);  
    CvMat* MatV = cvCreateMat(V->height,V->width,CV_64FC1);  
  
    CvMat* DCTY = cvCreateMat(Y->height, Y->width,CV_64FC1);  
    CvMat* DCTU = cvCreateMat(U->height, U->width,CV_64FC1);  
    CvMat* DCTV = cvCreateMat(V->height, V->width,CV_64FC1);  
      
    cvScale( Y, MatY );  
    cvScale( U, MatU );  
    cvScale( V, MatV );  
  
    cvDCT(MatY, DCTY, CV_DXT_FORWARD); //余弦变换   
    cvDCT(MatU, DCTU, CV_DXT_FORWARD); //余弦变换   
    cvDCT(MatV, DCTV, CV_DXT_FORWARD); //余弦变换   
      
    //Y 通道压缩   
    for(int i = 0; i < Y->height; i++)    
    {  
        for(int j = 0; j < Y->width; j++)  
        {  
            double  element = CV_MAT_ELEM( *DCTY, double, i, j );  
            if ( abs(element) < 10 )  
                CV_MAT_ELEM( *DCTY, double, i, j ) = 0;  
        }  
    }  
  
    // U 通道压缩   
    for(int i = 0; i < U->height; i++)    
    {  
        for(int j = 0; j < U->width; j++)  
        {  
            double  element = CV_MAT_ELEM( *DCTU, double, i, j );  
            if (     abs(element) < 20 )  
                CV_MAT_ELEM( *DCTU, double, i, j ) = 0;  
        }  
    }  
  
    // V 通道压缩   
    for(int i = 0; i < V->height; i++)    
    {  
        for(int j = 0; j < V->width; j++)  
        {  
            double  element = CV_MAT_ELEM( *DCTV, double, i, j );  
            if (     abs(element) < 20 )  
                CV_MAT_ELEM( *DCTV, double, i, j ) = 0;  
        }  
    }  
    cvDCT(DCTY, MatY, CV_DXT_INVERSE); //余弦反变换   
    cvDCT(DCTU, MatU, CV_DXT_INVERSE);   
    cvDCT(DCTV, MatV, CV_DXT_INVERSE);   
  
    cvScale( MatY, Y );  
    cvScale( MatU, U );  
    cvScale( MatV, V );  
      
    cvMerge( Y, U, V, NULL, YUVImage );  
    cvCvtColor( YUVImage, dst, CV_YUV2BGR); //YUV→BGR   
  
    //  计算前后两幅图像的PSNR值   
    for(int i = 0; i < src->height; i++)  
    {  
        for(int j = 0; j < src->width; j++)  
        {  
            SrcPixel = cvGet2D( src, i, j );  
            DstPixel = cvGet2D( dst, i, j );  
            sumB += ( SrcPixel.val[0] - DstPixel.val[0] ) * ( SrcPixel.val[0] - DstPixel.val[0] ); 
            sumG += ( SrcPixel.val[1] - DstPixel.val[1] ) * ( SrcPixel.val[1] - DstPixel.val[1] ); 
            sumR += ( SrcPixel.val[2] - DstPixel.val[2] ) * ( SrcPixel.val[2] - DstPixel.val[2] ); 
              
        }  
    }  
    mseB = sumB / ((src->width) * (src->height)); //计算均方差   
    mseG = sumG / ((src->width) * (src->height));   
    mseR = sumR / ((src->width) * (src->height));   
  
    PSNR_B = 10.0 * ( log10( 255.0 * 255.0 / mseB ) );  
    PSNR_G = 10.0 * ( log10( 255.0 * 255.0 / mseG ) );  
    PSNR_R = 10.0 * ( log10( 255.0 * 255.0 / mseR ) );  
    PSNR=(PSNR_B + PSNR_G + PSNR_R) / 3;  
      
    cvShowImage( "YImage", Y );  
    cvShowImage( "UImage", U );  
    cvShowImage( "VImage", V );  
    cvShowImage( "DstImage", dst );  
    cvSaveImage( "D:/demo/dstdemo.jpg", dst); 
  
    while( 1 )  
    {  
        if( cvWaitKey(0) == 27 ) break;  
    }  
  
    cvDestroyWindow("YImage");  
    cvDestroyWindow("UImage");  
    cvDestroyWindow("VImage");  
    cvDestroyWindow("DstImage");  
  
      
    cvReleaseImage(&Y);  
    cvReleaseImage(&U);  
    cvReleaseImage(&V);  
    cvReleaseImage(&src);  
    cvReleaseImage(&dst);  
    cvReleaseImage(&YUVImage);  
  
    return 0; 

}

3.参考用，jpg图片越压缩越大，代码有待优化来达到压缩效果！