How to compute P.S.N.R value for the below set of images?

I would request anybody to provide me with code to compute the PSNR value for the below pair of images. I want to compare output1.png with o.png, o2.png with output2.png? Please help?

Answer 1

[Thomas Koelen]

If you have the image processing toolbox you can use the function:

peaksnr = psnr(A,ref)

where A is your output1 and ref is o.

[Sandeep V]

A=imread('output1.png');
>> B=imread('o.png');
>> peaksnr = psnr(A,B);
??? Undefined function or method 'psnr' for input arguments of type 'uint8'.

I am using MATLAB R2010a version.

Can you please suggest what the problem is?

A=<540x960x3 uint8>
B=<540x960x3 uint8>

[Thomas Koelen]

Can you run this:

v = ver;
any(strcmp('Image Processing Toolbox', {v.Name}))

and post results? What version of matlab do you have?

Edit: just read that you have 2010a, I don't think peaksnr is present in that build.

I suggest you try one of the following:

http://www.mathworks.com/matlabcentral/fileexchange/29500-image-error-measurements/content/PSNR.m

or you could code it yourself:

X = double(X); 
Y = double(Y); 
m = sum((X(:)-Y(:)).^2) / prod(size(X)); 
psnr = 10*log10(255*255/m);

However, instead of assuming the peak is 255, you should use max(X(:)) as the peak.

[Sandeep V]

Thanks it is working and I am getting psnr as 28.241

Does this suggest that the second image(output1.png) is better than the first image(o.png)?

[Image Analyst]

Of course not. psnr is a measure of how close two images are to each other, not an overall assessment of which image is "better".

[Sandeep V]

OK then what does this 28.241 suggest? Can you please explain?