We can tell from your extraction of R, G, and B, that s is a 3 dimensional array. Your code line [m,n] = size(s) makes it look like a 2 dimensional array. There is meaning to that line, but it is not to extract just the first and second dimensions: when you use more than one output argument to size() and the number of output arguments is smaller than the number of dimensions, then the last output argument will be given a value which is the product of all the dimensions from there on.
Your nested "for" loops do not change any parameters in the call to evalfis(), so it is not clear why you are bothering to loop.
After your image resize, s will be 128 x 128 x 3 so [Ld ad bd] is going to be 128 x 384 . Is your fis set up to expect 128 input variables?
