To make motion estimation step spatial homogeneity

Question

we want to make algorithms

(average) Md=1/(N^2-K) ∑||(x,y)-S(x,y)||

where N is the size of the block, S(x, y) is the location of the selected pixel nearest to the pixel at location (x, y) and K is the number of selected pixels. Lower the value of μd and σ2 d, the more is the spatial homogeneity of the sampling lattice.

For K=9 I write a matlab code for this algorithm. I didn't find what's wrong with code

for i=2:1:a-1  
    for j=2:1:b-1
        S=blok(i-1:i+1;j-1:j+1);
        sum=sum+abs(blok(i,j)-S(i,j));
    end 
end
Md =double(sum/((a*b)-9));

Answer 1

input is block 16x16 block=16x16 (piece of image)

i found S(x,y) making kernel[1 1 1; 1 1 1; 1 1 1] convolution with block

for i=1:1:14
  for j=1:1:14

    for m=-1:1
        for n=-1:1
            S(i+1,j+1)=S(i+1,j+1)+kernel(m+2,n+2).*blok(i-m+1,j-m+1);
        end
    end
   % S=blok(i-1:i+1;j-1:j+1);

    total=total+abs(blok(i+1,j+1)-S(i+1,j+1));


   end 
 end

its seems work for now but ı am not sure.