How do I get the brightest pixels from this image?

The image in question is above. I want to get the pixel values of the bright ring that lines the center of this uneven ring.

This was a binary image (kind of an uneven donut), then I did the distance transform of the image. What you see is the result.

There is a ridge along the middle of the donut that contains the brightest pixels (farthest from edge). I want to algorithmically obtain all of those pixel values or the locations in the image where those pixel values are.

Suggestions are welcome. I do not need a fully coded solution... unless you want a coding challenge.

Answer 1

[Image Analyst]

You should be able to use @bwskel to get the skeleton. Then use the skeleton as a mask to get the pixel values along the ridge. See attached example where I get the mean width by examining along the spine of the distance transform. If you wanted the pixel intensities instead of the diameters, you'd do

skelImage = bwskel(mask); % Get skeleton of the binary image (not the distance transform).
% Get 1-D vector of all the pixel intensities in the original image along the mask spine.
pixelSkelValues = grayImage(skelImage); 

[Alexander Moody]

Excellent answer! Simple and very effective. Thank you!

I already tried it and it works really well. I forgot about bwskel() function because I never have a use for it.

Answer 2

[DGM]

See watershed()

% assuming D is a distance map
D = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1500965/image.png');
D = im2gray(D);

% find the ridgeline
mask = ~watershed(D);

% visualize both
imshow(imfuse(mask,mat2gray(D)),'border','tight')
xlim([450 1280]) % zoom in so it's visible in the forum
ylim([550 1350])

% get the distance at the ridgeline (the local width)
Dridge = D(mask);

For most tasks, the mask can be used directly instead of trying to use subscripts. If you need subscripts for geometric purposes, use find().

This example has a different end goal, but it demonstrates the process of addressing image pixels using three different appcoaches: a mask, linear indices derived from the mask, and row/column subscripts derived from the mask.

https://www.mathworks.com/matlabcentral/answers/2024752-extract-part-from-an-rgb-image-based-on-the-obtained-pixel-locations#comment_2898297

That should serve to substantiate my recommendation to use the mask alone for addressing tasks.

[Alexander Moody]

You know, I was thinking about using a watershed, but the problem is that it doesn't capture the whole ring. This is a great answer though. I really appreciate your help. Thank you!