您的位置:首页 > 其它

图像的 SNR 和 PSNR 的计算

2016-03-31 10:39 615 查看
PSNR 的公式很容易搜到。
http://www.360doc.com/content/12/0605/21/4129998_216244993.shtml http://blog.sina.com.cn/s/blog_455c7a600101ytgo.html

峰值信噪比经常用作图像压缩等领域中信号重建质量的测量方法,它常简单地通过均方差(MSE)进行定义。两个m×n单色图像I和K,如果一个为另外一个的噪声近似,那么它们的的均方差定义为:



峰值信噪比定义为:



代码实现(参考:http://stackoverflow.com/questions/29428308/snr-of-an-image-in-c-using-opencv)

double getPSNR(const Mat& I1, const Mat& I2)
{
Mat s1;
absdiff(I1, I2, s1);       // |I1 - I2|
s1.convertTo(s1, CV_32F);  // cannot make a square on 8 bits
s1 = s1.mul(s1);           // |I1 - I2|^2

Scalar s = sum(s1);         // sum elements per channel

double sse = s.val[0] + s.val[1] + s.val[2]; // sum channels

if( sse <= 1e-10) // for small values return zero
return 0;
else
{
double  mse =sse /(double)(I1.channels() * I1.total());
double psnr = 10.0*log10((255*255)/mse);
return psnr;
}
}


SNR 不太好搜。
http://cg2010studio.com/2014/12/10/opencv-snr-%E8%88%87-psnr/ http://blog.csdn.net/lien0906/article/details/30059747
SNR (Signal to Noise Ratio):訊號雜訊比,簡稱訊雜比。



PSNR (Peak Signal to Noise Ratio):也是訊雜比,只是訊號部分的值通通改用該訊號度量的最大值。以訊號度量範圍為 0 到 255 當作例子來計算 PSNR 時,訊號部分均當成是其能夠度量的最大值,也就是 255,而不是原來的訊號。





代码实现(参考:http://cg2010studio.com/2014/12/10/opencv-snr-%E8%88%87-psnr/)

/**
Theme: SNR (Signal to Noise Ratio) & PSNR (Peak Signal to Noise Ratio)
compiler: Dev C++ 4.9.9.2
Library: OpenCV 2.0
Date: 103/12/10
Author: HappyMan
Blog: https://cg2010studio.wordpress.com/ */
#include <cv.h>
#include <highgui.h>
#include<iostream>

using namespace std;

int main(){
IplImage *src1= cvLoadImage("moon_o.BMP");
IplImage *src2= cvLoadImage("moon_m.BMP");

long long int sigma = 0;
long long int squre = 0;
double MSE = 0.0;
double SNR = 0.0;
double PSNR = 0.0;
int frameSize = src1->height*src1->width*3;
int blue1=0, blue2=0;
int green1=0, green2=0;
int red1=0, red2=0;

// width x height -> [height][width]
for(int i=0;i<src1->height;i++){
for(int j=0;j<src1->widthStep;j=j+3){
blue1=(int)(uchar)src1->imageData[i*src1->widthStep+j];//Blue
green1=(int)(uchar)src1->imageData[i*src1->widthStep+j+1];//Green
red1=(int)(uchar)src1->imageData[i*src1->widthStep+j+2];//Red
blue2=(int)(uchar)src2->imageData[i*src2->widthStep+j];//Blue
green2=(int)(uchar)src2->imageData[i*src2->widthStep+j+1];//Green
red2=(int)(uchar)src2->imageData[i*src2->widthStep+j+2];//Red
sigma+=(blue1-blue2)*(blue1-blue2)+
(green1-green2)*(green1-green2)+
(red1-red2)*(red1-red2);
squre += blue1*blue1 + green1*green1 + red1*red1;
}
}
MSE=sigma/(double)frameSize;
PSNR=10*log10(255*255/MSE);
SNR = 10*log10(squre/sigma);

cout<<"sigma: "<<sigma<<endl;;
cout<<"MSE: "<<MSE<<endl;;
cout<<"PSNR: "<<PSNR<<endl;;
cout<<"SNR: "<<SNR<<endl;;

system("pause");
cvWaitKey(0);
return EXIT_SUCCESS;
}
内容来自用户分享和网络整理,不保证内容的准确性,如有侵权内容,可联系管理员处理 点击这里给我发消息
标签: