extract blocks using the Mean Square Error matlab - matlab

Question

The Mean Square Error(MSE), is a method used to define the difference in between two blocks, and can be calculated as follow: a and b two blocks equal size
MSE = sqrt(sum(sum((a-b).^2)))/size(a or b)
If the MSE is less than a given threshold, than the two blocks are not different.
Given a matrix A, already reshaped to be contain blocks all in the same raw,
the purpose is to extract all blocks where the MSE is less than a given threshold (based on the first block), then return the mean of those blocks. again, extract the second group of blocks which the MSE is less than the given threshold where the blocks that already assigned to be a part of other group of blocks must not be extracted again. Better than that, it must be deleted to reduce the search time. and so on till all blocks of the matrix A are assigned to be part of a group. And here is an example :
Given matrix A where the size of A is 2 by 14:
A= [1 1 2 2 9 9 4 4 6 6 5 5 3 3
1 1 2 2 9 9 4 4 6 6 5 5 3 3];
PS: its not necessary the blocks contain the same numbers, it is just to make the example clear.
blocks size is : 2 by 2
the threshold is 2
now we extract all blocks where the MSE is less than the threshold starting from the first block in the matrix A. so the blocks are:
1 1 2 2 3 3
1 1 2 2 3 3
the mean of those blocks is
Result= [ 2 2
2 2];
again. we extract all blocks where the MSE is less than the threshold, but we need to avoid the blocks that already extracted, so the second group of blocks is :
9 9
9 9
the mean of this block is it self, so:
Result= [2 2 9 9
2 2 9 9];
again. we extract all blocks where the MSE is less than the threshold, but we need to avoid the blocks that already extracted, so the third group of blocks is :
4 4 6 6 5 5
4 4 6 6 5 5
the block
3 3
3 3
is not a part of this group even if the MSE is less then the threshold because is already extracted to be part of the first group.
the mean of those blocks is:
5 5
5 5
therefore the result should be:
Result= [2 2 9 9 5 5
2 2 9 9 5 5];
there are any fast way to apply that ?

Answer 1

Here's my solution:
A = [1 1 2 2 9 9 4 4 6 6 5 5 3 3
1 1 2 2 9 9 4 4 6 6 5 5 3 3];
blockLen = 2;
MSEthresh = 1;
nBlocks = size(A,2) / blockLen;
meanBlocks = [];
bSelectRefBlock = true;
while(~isempty(A))
if(bSelectRefBlock)
%// Select new reference block
refBlock = A(:, 1:blockLen);
bSelectRefBlock = false;
A = A(:, blockLen+1:end);
iBlock = 1;
selectedBlocks = refBlock;
else
if(iBlock > size(A,2))
%// End of A reached. Compute mean of selected blocks
blockCols = reshape(selectedBlocks, [numel(refBlock) numel(selectedBlocks)/numel(refBlock)]);
meanBlockCols = mean(blockCols, 2);
meanBlocks(:, end+1:end+blockLen) = reshape(meanBlockCols, [size(refBlock, 1) blockLen]);
bSelectRefBlock = true;
else
%// Test next block with MSE and select if below threshold
MSE = sqrt(sum(sum((A(:, iBlock:(iBlock+blockLen-1)) - refBlock).^2))) / numel(refBlock);
if(MSE <= MSEthresh)
selectedBlocks(:, end+1:end+blockLen) = A(:, iBlock:(iBlock+blockLen-1));
A(:, iBlock:(iBlock+blockLen-1)) = [];
else
iBlock = iBlock + blockLen;
end
end
end
end
%// Compute mean of selected blocks for last group
blockCols = reshape(selectedBlocks, [numel(refBlock) numel(selectedBlocks)/numel(refBlock)]);
meanBlockCols = mean(blockCols, 2);
meanBlocks(:, end+1:end+blockLen) = reshape(meanBlockCols, [size(refBlock, 1) blockLen]);