[SOLVED] conflict about the SNR in image

[Serwan Bamerni]

Hi everyone

I am trying to add noise of 5dB to an image in matlab according to the following steps

SNR = 5;
Im = double(Io) / 255;
vm = var(Im(:)) / 10^(SNR/10);
I_my_noisy = imnoise(Im, 'gaussian', 0, vm);

where Io is the readed original image

but when I measure the SNR by using the matlab command

[peaksnr,snr] = psnr(I_my_noisy, Im)

I get SNR equal 10. So, what is the wrong in my code? and why the result differ from my add dB?

 

[Serwan Bamerni]

Hi everyone

I went deeply in calculating the SNR for image in my case and I found that the add SNR is obtained via the following code

Error = I_my_noisy - Im;
SNR = 10*log10(var(Im(:))/var(Error(:)));

and that differ from the result obtained from the direct matlab function

[peaksnr,snr] = psnr(I_my_noisy, Im)

that calculate the power ratio between signal and noise

 

[c_mitra]

I presume the noise is being measured as power and is being defined as 20 log xxx.

You need to modify your code:

vm = var(Im(:)) / 10^(SNR/10);

and try again and see. I do not see any conflict.

 

[Serwan Bamerni]

I presume the noise is being measured as power and is being defined as 20 log xxx.

You need to modify your code:

vm = var(Im(:)) / 10^(SNR/10);

and try again and see. I do not see any conflict.

Hello Mr. mitra

I depend on the definition of the SNR as in this link
http://bigwww.epfl.ch/sage/soft/snr/

 

[c_mitra]

Please check the help menu on psnr. I do not have matlab and I use octave.