Find Max Delta of Sliding 3x3 Window Over Matrix
显示 更早的评论
I have the code below to find the max delta between a point and its 8 neighbors in a sliding 3x3 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.
采纳的回答
Hi Paul,
You can use the MATLAB ‘conv2’ function to compute the Max-Delta of Sliding 3x3 Window by subtracting the center pixel with all 8 neighbours and taking the absolute difference.
Using an appropriate kernel can be helpful while dealing with the ‘conv2’ function, the following kernels can be used for the calculation:
% Kernels of neighbours are as follows: Right, Left, Up, Down, Top-Left, Top-Right, Bottom-Left, Bottom-Right
kernels = {
[0 0 0; 0 -1 1; 0 0 0],
[0 0 0; 1 -1 0; 0 0 0],
[0 1 0; 0 -1 0; 0 0 0],
[0 0 0; 0 -1 0; 0 1 0],
[1 0 0; 0 -1 0; 0 0 0],
[0 0 1; 0 -1 0; 0 0 0],
[0 0 0; 0 -1 0; 1 0 0],
[0 0 0; 0 -1 0; 0 0 1]
};
To handle edge cases and ensure accurate convolution, you can add the padding to the input matrix as follows:
A = magic(3)
padPt = mean([min(A(:)), max(A(:))]);
% Pad the matrix with mean value
Apad = padarray(A, [1, 1], padPt, 'both');
Refer to the following example code snippet to apply Convolution to compute differences and obtain Max-Delta result:
diffs = zeros([size(Apad), 8]);
% Apply each kernel with conv2 and store the absolute difference
for k = 1:8
diffs(:,:,k) = abs(conv2(Apad, kernels{k}, 'same'));
end
% Using the maximum difference across all 8 neighbors
Ctmp = max(diffs, [], 3);
% Output after removing padding
res = Ctmp(2:end-1, 2:end-1)
Refer to the following MathWorks Documentation to learn more about ‘conv2’ function: https://www.mathworks.com/help/releases/R2024a/matlab/ref/conv2.html
I hope this helps you in resolving your query.
更多回答（0 个）
类别
在 帮助中心 和 File Exchange 中查找有关 Image Processing Toolbox 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
