i have a greyscale image of size 512x512 and i have divided into 4x4 non overlapping blocks so i get 16384 blocks to process, by processing it i get the last blocks only.

my code is

X=mat2cell(I1,4*ones(1,128),4*ones(1,128));
          for i=1:16384
          A = X{i}(:);
          Di = abs(diff(A));
          end

here i got the all the block but when the loops end and processing is finish it return only the last block.

[KSSV]

Why you think you got only last block?

[John Chilleri]

Di = abs(diff(A));

Did you mean:

D(i) = abs(diff(A));?

Hope this helps!

Answer 1

[Guillaume]

Well, in the expression Di = ... matlab doesn't know that you want to replace the i by whichever value i is at the moment. So obviously, you're overwriting the variable Di at each step of the loop. The fix is to use indexing: D(i) = .... To avoid resizing the destination array each time you add an element, you should predeclare it before the loop with D = zeros(numel(X)), resulting in:

X = mat2cell(I1, 4*ones(1, 128), 4*ones(1, 128));
D = zeros(numel(X));
for i = 1:numel(X)  %don't hardcode sizes when you can just query them
     A = X{I}(:);
     D(i) = abs(diff(A));
end

Or... you could use blockproc which does the separating into block, looping and creating the destination array all for you:

D = blockproc(I1, [4, 4], @(block) abs(diff(block.data(:))));

[KSSV]

I1 = rand(512,512) ;
X=mat2cell(I1,4*ones(1,128),4*ones(1,128));
Di = zeros(numel(X),15) ;
for i=1:16384
    A = X{i}(:);
    Di(i,:) = abs(diff(A));
end

[Guillaume]

Oh, yes of course, the result of the diff is a vector. Thanks for the correction.

As said, using blockproc would mean not having to worry about creating a destination the correct size.