Line 2 creates an anonymous function that calculates the median of a given numeric array. It's what will process the image content within the moving window.
You can use nlfilter() if you want, but medfilt2() would probably make more sense. Still, that's just a plain median filter, not really a noise removal filter with any sort of discrimination. As the filter size gets larger or the noise density increases, the difference becomes notable.
inpict = imread('circuit.tif');
inpict = imnoise(inpict,'salt & pepper',0.2);
imshow(inpict)
A = medfilt2(inpict,[3 3]); % simple local median without noise discrimination
imshow(A) % it doesn't really get rid of noise due to the small window size
B = medfilt2(inpict,[7 7]); % so try a larger window size
imshow(B) % but that destroys details
C = amedfiltsimple(inpict); % adaptive median noise removal filter
imshow(C) % only filter the locations where there's noise
