I have the code below to find the max delta between a point and its 8 neighbors in a sliding 3×3 window. I do this for the whole matrix input, A. This works fine and correct, but I am wondering if this can be accomplished with several uses of conv2 instead of the imerode/imdilation calls. The computation I’m doing is depicted in the attached image.
% This method works fine…
A = magic(3); % Test case
% This finds the max diff
% between a pixel and its 8 neighbors.
drMin = min(A(:)); drMax = max(A(:)); % Phantom (padding) pts outside of domain are assumed the avg.
padPt = mean([drMin drMax]); % Value to use for padding.
tmp1 = padarray(A,1,padPt,’both’); tmp2 = padarray(tmp1′,1,padPt,’both’); Apad = tmp2′; % Pad matrix.
maxApad = imdilate(Apad,ones(3)) – Apad; % Local max minus the map. Delta from each pt to max neighbor.
minApad = imerode(Apad,ones(3)) – Apad; % Local min minus the map. Delta from each pt to min neighbor.
tmp(:,:,1) = abs(maxApad); tmp(:,:,2) = abs(minApad);
Ctmp = max(tmp,[],3); % Largest of the two gives the max delta from each point to its neighbors!
C = Ctmp(2:end-1,2:end-1); % Remove padding…

The expected result for the magic(3) input is:
7 7 5
6 4 6
5 7 7

I have tried with the following code, but it’s clearly off.

A = magic(3);

arrayTot = zeros(3,3,8);
kernel = zeros([9 1]); kernel(5) = 1;

for k = 1:9

if k ~=5
kernel(k) = -1;
k1 = kernel;
k1 = reshape(k1,[3 3]);
tmp = conv2(A,k1,’same’);
arrayTot(:,:,k) = abs(tmp);
kernel = zeros([9 1]); kernel(5) = 1;
end
end
tst = max(arrayTot,[],3)
Any advice would be appreciated. Thanks.I have the code below to find the max delta between a point and its 8 neighbors in a sliding 3×3 window. I do this for the whole matrix input, A. This works fine and correct, but I am wondering if this can be accomplished with several uses of conv2 instead of the imerode/imdilation calls. The computation I’m doing is depicted in the attached image.
% This method works fine…
A = magic(3); % Test case
% This finds the max diff
% between a pixel and its 8 neighbors.
drMin = min(A(:)); drMax = max(A(:)); % Phantom (padding) pts outside of domain are assumed the avg.
padPt = mean([drMin drMax]); % Value to use for padding.
tmp1 = padarray(A,1,padPt,’both’); tmp2 = padarray(tmp1′,1,padPt,’both’); Apad = tmp2′; % Pad matrix.
maxApad = imdilate(Apad,ones(3)) – Apad; % Local max minus the map. Delta from each pt to max neighbor.
minApad = imerode(Apad,ones(3)) – Apad; % Local min minus the map. Delta from each pt to min neighbor.
tmp(:,:,1) = abs(maxApad); tmp(:,:,2) = abs(minApad);
Ctmp = max(tmp,[],3); % Largest of the two gives the max delta from each point to its neighbors!
C = Ctmp(2:end-1,2:end-1); % Remove padding…

The expected result for the magic(3) input is:
7 7 5
6 4 6
5 7 7

I have tried with the following code, but it’s clearly off.

A = magic(3);

arrayTot = zeros(3,3,8);
kernel = zeros([9 1]); kernel(5) = 1;

for k = 1:9

if k ~=5
kernel(k) = -1;
k1 = kernel;
k1 = reshape(k1,[3 3]);
tmp = conv2(A,k1,’same’);
arrayTot(:,:,k) = abs(tmp);
kernel = zeros([9 1]); kernel(5) = 1;
end
end
tst = max(arrayTot,[],3)
Any advice would be appreciated. Thanks. I have the code below to find the max delta between a point and its 8 neighbors in a sliding 3×3 window. I do this for the whole matrix input, A. This works fine and correct, but I am wondering if this can be accomplished with several uses of conv2 instead of the imerode/imdilation calls. The computation I’m doing is depicted in the attached image.
% This method works fine…
A = magic(3); % Test case
% This finds the max diff
% between a pixel and its 8 neighbors.
drMin = min(A(:)); drMax = max(A(:)); % Phantom (padding) pts outside of domain are assumed the avg.
padPt = mean([drMin drMax]); % Value to use for padding.
tmp1 = padarray(A,1,padPt,’both’); tmp2 = padarray(tmp1′,1,padPt,’both’); Apad = tmp2′; % Pad matrix.
maxApad = imdilate(Apad,ones(3)) – Apad; % Local max minus the map. Delta from each pt to max neighbor.
minApad = imerode(Apad,ones(3)) – Apad; % Local min minus the map. Delta from each pt to min neighbor.
tmp(:,:,1) = abs(maxApad); tmp(:,:,2) = abs(minApad);
Ctmp = max(tmp,[],3); % Largest of the two gives the max delta from each point to its neighbors!
C = Ctmp(2:end-1,2:end-1); % Remove padding…

The expected result for the magic(3) input is:
7 7 5
6 4 6
5 7 7

I have tried with the following code, but it’s clearly off.

A = magic(3);

arrayTot = zeros(3,3,8);
kernel = zeros([9 1]); kernel(5) = 1;

for k = 1:9

if k ~=5
kernel(k) = -1;
k1 = kernel;
k1 = reshape(k1,[3 3]);
tmp = conv2(A,k1,’same’);
arrayTot(:,:,k) = abs(tmp);
kernel = zeros([9 1]); kernel(5) = 1;
end
end
tst = max(arrayTot,[],3)
Any advice would be appreciated. Thanks. convolution, image processing, imdilate, imerode MATLAB Answers — New Questions

​