Finding the neighbouring regions

How I can find the neighbouring regions of each region or how to find the region sharing commn boundary

Answer 1

[Image Analyst]

Not too hard. Simply call bwlabel to give each region an ID. Then call bwmorph with the thicken option to dilate the regions to get rid of the black border between regions. Then use the gray level cooccurrence matrix, graycomatrix, to discover which regions are next to each other. I'm attaching a demo of graycomatrix. Adapt as needed.

[Image Analyst]

You said that @Muhammad Hassaan Shah suggestion did not work for you. Strange because it's essentially what I suggested first, except that scaling he did shouldn't be necessary. So attach your original binary image (not annotated with region numbers) and I'll see if tomorrow I have time to give you a solution that works.

[Akhil]

I am sharing the original binary image. Thanks in advance for the help

[Image Analyst]

Attach the image file, not a fig file. So the PNG, TIFF, JPG, or BMP format image. Or the variable in a .mat file.

[Akhil]

I am sharing the JPG file.

[Image Analyst]

It's not working. The JPG corruption and artifacts are cluttering up the black dividing lines. Can you save it as PNG format?

[Akhil]

Kindly find the image in the attachment.

Answer 2

[Hassaan]

% Assuming 'binaryImage' is your binary segmented image, where each object
% is separated by black borders.

% Step 1: Label each region with a unique identifier using bwlabel.
[labeledImage, numRegions] = bwlabel(binaryImage);

% Step 2: Thicken the regions to remove black borders using bwmorph.
% This step may not be necessary if your regions are already adjacent to each other.
% Adjust the number of iterations of thickening as needed.
thickenedImage = bwmorph(labeledImage, 'thicken', 1);

% Step 3: Use graycomatrix to find adjacent regions.
% This step involves creating a gray level co-occurrence matrix for the labeled image.
% For this, you might need to loop over each pair of regions to check adjacency.
% Since graycomatrix works on grayscale images, we ensure our labeled image is in the
% range 0-255 (typical for grayscale images).
scaledLabeledImage = uint8(255 * mat2gray(labeledImage));

% Initialize adjacency matrix
adjacencyMatrix = zeros(numRegions);

% Find adjacent regions
for region1 = 1:numRegions
    for region2 = region1+1:numRegions
        % Create a binary image for each pair of regions
        binaryPair = ismember(scaledLabeledImage, [region1, region2]);
        
        % Calculate the co-occurrence matrix for the pair of regions
        glcm = graycomatrix(binaryPair, 'Offset', [0 1; -1 0; -1 1; -1 -1]);
        
        % If there is co-occurrence, the regions are adjacent
        if any(glcm(:) > 1)
            adjacencyMatrix(region1, region2) = 1;
            adjacencyMatrix(region2, region1) = 1;
        end
    end
end

% Now, adjacencyMatrix contains information about which regions are adjacent.
% An entry of 1 at (i,j) means region i is adjacent to region j.

% Please note that the exact implementation may vary depending on your specific
% image and requirements. This is a general approach and may require adaptations.

Make sure to replace binaryImage with your actual segmented image variable. The code above assumes you have a binary image where objects are marked with ones and the background is marked with zeros. The bwlabel function labels connected objects with unique identifiers. The bwmorph function with 'thicken' is used to ensure that adjacent regions touch each other, which may or may not be necessary depending on your specific image. The graycomatrix function is used to create a gray level co-occurrence matrix, which is then used to determine if regions are adjacent.

You'll need to test and possibly adjust the code to suit your specific image and analysis needs.

Note

This is a general approach and may require adaptations.

------------------------------------------------------------------------------------------------------------------------------------------------

If you find the solution helpful and it resolves your issue, it would be greatly appreciated if you could accept the answer. Also, leaving an upvote and a comment are also wonderful ways to provide feedback.

Professional Interests

• Technical Services and Consulting
• Embedded Systems
• Electrical and Electronics Engineering

[Akhil]

Thanks for the help. I tried the suggested code and made some of my adjustments, however, adjacencyMatrix giving values 1 where i is not equal to and 0 where i and j equal to zero.