MMSE of a Gaussian noisy image

Hi everyone,

I have a Gaussian noise corrupted image and I need to know how can I find the MMSE estimate of the entire noisy image in MATLAB? or how can I find the MMSE estimate of a Gaussian noise vector/signal (because I can divide the entire image into small patch vectors)?

Thanks.

Answer 1

[Image Analyst]

The Image Processing Toolbox has a function immse(). Use that.

[Muzammil Behzad]

Thank you for your response.

Unfortunately I cannot see the immse() function in my MATLAB toolbox. Can you please tell me how can I get it or can you please give me a link where I can download the source code of this function? Also, is there any documentation of the mathematical approach this function is using?

Thanks.

[Image Analyst]

It was introduced in R2014b so you should probably have it. If you have an obsolete version of MATLAB, then see my attached demo.

[Muzammil Behzad]

Thank you for your response.

The file you attached is about finding psnr whereas I am asking about finding the MMSE estimate of a noisy image. I didn't get the relationship of finding the estimate (denoised image) via immse() with psnr(). Can you please send me the immse function file? I'll be grateful. Thanks.

[Image Analyst]

I can't send you immse() (just one file out of a more modern toolbox) - you'd need to upgrade your MATLAB.

[Muzammil Behzad]

Ok thanks for your help.

I'll try to upgrade and will then see if the problem gets solved. Thanks.

[Muzammil Behzad]

Hi again.

I upgraded to MATLAB 2014b but when I checked the immse() function I came to notice that you may haven't gotten my question. My question is that I have a noisy image and I want to find its estimate/denoised version by using MMSE (minimum mean square error) as estimator, so how can I do it in MATLAB? What approach/steps should I use to get the denoised version using MMSE estimator?

Thanks.