PSNR- Peak Signal to Noise Ratio

compute the Peak Signal to Noise Ratio of two images

Description

defined by :
PSNR(x,y) = 10*log10( max(max(x),max(y))^2 / |x-y|^2 ).

• MATLAB release: MATLAB 7.7 (R2008b)

[Adi Navve]

This does not conform with the standard definition of PSNR.

The signal in the calculation needs to be the maximum possible signal, for example 255 for 8bit images. In your case, with dark images the value would be much too low.

For example:

img1 = double(rgb2gray(imread('lena.jpg')));
img1 = uint8(img1/4);
img2 = imnoise(img1,'gaussian',0,0.0001);

figure
subplot(1,2,1)
imshow(img1)
title('original')
subplot(1,2,2)
imshow(img2)
title('minimal noise added')
psnr = psnr(img1,img2)

psnr = 28.0867

while the value should actually be 39.9592