Getting high PSNR on 16 bit DICOM MRI. What am i doing wrong?

  clc;clear all;close all;
  info=dicominfo('IMG_000001_original.dcm');
  A=dicomread(info);
  info=dicominfo('IMG_000001_(im.std =5).dcm');
  B=dicomread(info);
  [M, N ] = size(A);
  ms= sum(sum((double(A)-double(B)).^2));
  mse=ms/(N*M);
  Y=immse(A,B);
  P=10*log10((65535^2)/mse); 
  K=psnr(A,B); 
  Result = sprintf('The mean square error = %.2f %.2f',mse, P)

I am denoising MR images. A is the original image. B is the denoised image. Both are 16-bit DICOM images. When I try to find the PSNR between them, im getting very high values, like 90db, 100db ,etc when im varying standard deviation of gaussian noise as 2,10, 50, 100 etc. I read somewhere that 6N+12 db is the maximum PSNR you can have for an N-bit image so for a 16-bit image, that would be 108 dB. But i cant find any research papers with such high values so im probably doing something wrong.

I tried converting the images to png and tiff but got the same results. I also used uint8 to convert the images to 8 bit but there is loss of data. I have attached the workspace variables A and B for reference.

So what am i doing wrong here in calculating PSNR of 16 bit MR images.

[DGM]

Under what conditions would there be a finite limit on PSNR? If PSNR is finite, then that means error is nonzero. It's trivial to create an exact copy of an image, but I don't know what effects of your processing might limit the fidelity.

[anagha]

Wait im not quite following. Should i just continue with the values im getting? And is there anything i could change in the code to arrive at a different result?

[DGM]

Like I said, I don't know why you are expecting there to be a lower bound on the noise. Without more information, I can't know if that particular limiting value is correct.

All I know is that your manual calculation of PSNR follows from the same method used by the native psnr() function. That much doesn't answer your question about the process which is governing noise presence.

Maybe you're thinking about the limitations on SNR when capturing images? If so, this is probably best answered by someone who is more familiar with the subject. (Hence the fact that I'm only responding in comments)