distance between an image and a sliding window

Hi ! I have an image I of size 300x300 pix, and a "filter" F of size 12x12 pix, and I would like to slide this patch everywhere on I and for each position, to extract the euclidean distance between F and the 12x12 local patch of I. Is there an efficient way to encode this ? In advance, Thanks !

[Rik]

Do you mean a convolution? If so, Matlab has a function to do that. I don't really understand what you mean with the distance between the sample and the filter. Could you provide a small example of how this should work?

Also, have a read here and here. It will greatly improve your chances of getting an answer.

[Julien Renoult]

Hi. Thanks for your comment. A kind of convolution but I do not want the product of the filter and the patch, but the euclidean distance. Currently I have this peace of code. My problem is that it is really time-consuming (I repeat this on thousands of images) and I thus would like to avoid these two "for" loops.

F = rand(12,12,4); % I did not mentioned it before but in fact I have 3D filters and images
I = rand(70,100,4);
sz = size(F,1);
[dx2,dy2,~]=size(I);
for y=1:dy2-sz
    for x=1:dx2-sz
        P = I(x:x+sz-1,y:y+sz-1,:);
        R = F - P;
        R = R.^2;
        R = sum(sum(sum(R)));
        J(x,y) = R;
     end
 end

Answer 1

[Image Analyst]

Convolve I with F first using convn(). Then use immse().