How can I create a sliding window centered on a current pixel to impose over an impulse noise affected gray-scale image?

How can I create a sliding window of size 11x11 centered on a current pixel to impose over a random-valued impulse noise affected image  of size 256x256 in MATLAB? I need to impose the window for every pixel starting from position (1,1) to (255,255). Therefore I have created a padded image of size (266,266) having the original noisy image in the middle position. Am I proceeding in the right direction? I want to find the maximum and minimum values within the window for each iteration.

Answer 1

[Guillaume]

The normal way of removing impulse noise is to deconvolve the image with the psf of the impulse. There's a function in matlab that can do that for you and also make a guess at the psf: deconvblind. If you want to do the deconvolution manually, you normally take the FFT of the image and of the psf. divide the two and take the inverse FFT of the result.

To find the maximum of the image over a sliding window you could do it with a loop and then, yes, creating the padded image is the way to do it. Or you could use the built-in nlfilter which may be faster than the loop and certainly a lot less code

maxfiltered = nlfilter(img, [11 11], @(b) max(b(:)));
minfiltered = nlfilter(img, [11 11], @(b) min(b(:)));

The downside of nlfilter is that it loops over the image twice to get min and max, whereas with the loop you can do both at once in the same loop.

[Mayuri]

what does the variable 'b' represent in the function handle?

[Guillaume]

b is your 11x11 block, zero-padded for the edges.