Why do I receive "attempt to grow array along ambiguous dimension" when trying to apply a binary mask over a greyscale image?

I am working on a particle detection project and I am using the Hough method to detect the particles (white rings on a greyscale background). Once the particle is detected I want to crop it and apply a black circular mask over it in order to exclude overlapping particles from my dataset. (If the detected particle overlaps with another particle the intensity outside of the masked area is above my threshold and the image is rejected). The code works fine for about 70-100 iterations and after that I get the following error message:

Attempt to grow array along ambiguous dimension.

Error in F1_save_PI_M2 (line 31)

greyscale(circleMask) = 0;

Here is the Code:

for i=1:length(xA)
        D = dA(i);
        if(D > 0)
            %% Cutting out particle images
            pIN = pIN +1; %running variable for particle image number
            ol = 63; %outline around particle center
            lb = max(yA(i)-ol,1);
            rb = min(yA(i)+ol+1,size(A,1));
            bb = max(xA(i)-ol,1);
            ub = min(xA(i)+ol+1,size(A,2));
            partIm = A(lb:rb, bb:ub); %stiff image size definition
       
            
            %% Reject Overlaps, random noise by masking image
            % Create circular mask
            imageSizeX = 128;
            imageSizeY = 128;
            cX = 63;
            cY = 63;
            R = D/2 + ovl;
            [columns rows] = meshgrid(1:imageSizeX, 1:imageSizeY);          
            circleMask = (rows - cY).^2 + (columns - cX).^2 <= R.^2;
            greyscale=partIm;
          
            greyscale(circleMask) = 0;

Answer 1

[Walter Roberson]

partIm = A(lb:rb, bb:ub)

That would typically be a 2D array.

            imageSizeX = 128;
            imageSizeY = 128;
            % stuff
            [columns rows] = meshgrid(1:imageSizeX, 1:imageSizeY);          
            circleMask = (rows - cY).^2 + (columns - cX).^2 <= R.^2;

That is a logical array that is 128 x 128, which is bigger than partIm

            greyscale=partIm;
            greyscale(circleMask) = 0;

circleMask is larger than grayscale, so assigning 0 into those locations would requiring expanding the array. MATLAB is confused about how to do that growing.

You could do

            if size(grayscale,1) < imageSizeY; grayscale(imageSizeY,1) = 0; end
            if size(grayscale,2) < imageSizeX; grayscale(1,imageSizeX) = 0; end
            

Those lines would, if necessary, expand grayscale to the same size as circleMask.

Or... you could use

            imageSizeX = min(128, size(partIm,2));
            imageSizeY = min(128, size(partIm,1));
            % stuff
            [columns rows] = meshgrid(1:imageSizeX, 1:imageSizeY);          
            circleMask = (rows - cY).^2 + (columns - cX).^2 <= R.^2;

and then you would know that circleMask would be no larger than partIm.

I have to wonder, though, whether you really want to apply the circle mask to whatever whatever the selected portion of the image is as if that selected window was the upper-left corner of the circle. It would seem to me to make more sense to create a circle mask the size of the original image, and then extract the same portion of it as you extracted for partIm, and mask.

[Radoslav Yankov]

Thank you very much for your response, Walter. The script works now. I can discuss the last paragraph of your proposal with my supervisor as most of the code for the project is his.