How can I "get rid" of certain regions in my segmented image based on their dimensions?

Question

Andrea Labudzki 2022-2-3

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1642190-how-can-i-get-rid-of-certain-regions-in-my-segmented-image-based-on-their-dimensions

评论： Andrea Labudzki 2022-2-4

filteredImage.bmp

Hello everyone. I have a code that takes in a binarized image, segments it, and finds the most common range of width and height of the segments. I would like to now go through the regions and "delete" the regions whose dimensions do not fall within these most common ranges (so turn these segments' pixels from 1 to 0).

The image that I am using for the code is attached, and the code is below. I would appreciate any advice on this!

%%----------Segmentation----------------------------------------------
%img = (incoming image)
% Label each region with 8-connectivity
[labeledImage, numOfBlobs] = bwlabel(img, 8);
% Add pseudo-colors to blobs and create image
pseudoColoredImage = label2rgb (labeledImage, 'hsv', 'k', 'shuffle');
% Display pseudo-colored image
subplot(2, 3, 1);
imshow(pseudoColoredImage);
impixelinfo;
axis('on', 'image');
title('Pseudo-colored Image', 'FontSize', fontSize, 'Interpreter', 'None');
hold on
drawnow;
% Retrieve blob properties
blobProps = regionprops(labeledImage, img, 'Area', 'BoundingBox', 'Centroid'); 
numOfBlobs = size(blobProps, 1);
% Get bounding boxes
bounds = vertcat(blobProps.BoundingBox); %concatenates matrices together
widths = bounds(:, 3);
heights = bounds(:, 4);
% Get a list of all the dimensions 
allWidths = [widths];
allHeights = [heights];  
% Calculate mode dimensions
modeWidth = mode(allWidths, 'all');
modeHeight = mode(allHeights, 'all');
% Show the histograms of widths and heights
subplot(2, 3, 2);
histogram(allWidths);
xline(modeWidth, 'Color', 'r', 'LineWidth', 3)
caption = sprintf('Distribution of %d Widths.\nMode Width = %.2f pixels.', numOfBlobs, modeWidth);
title(caption, 'FontSize', fontSize, 'Interpreter', 'None');
xlabel('Width', 'FontSize', fontSize, 'Interpreter', 'None');
ylabel('Count', 'FontSize', fontSize, 'Interpreter', 'None');
grid on;
drawnow;
subplot(2, 3, 3);
histogram(allHeights);
xline(modeHeight, 'Color', 'r', 'LineWidth', 3)
caption = sprintf('Distribution of %d Heights.\nMode Height = %.2f pixels.', numOfBlobs, modeHeight);
title(caption, 'FontSize', fontSize, 'Interpreter', 'None');
xlabel('Height', 'FontSize', fontSize, 'Interpreter', 'None');
ylabel('Count', 'FontSize', fontSize, 'Interpreter', 'None');
grid on;
drawnow;
%-------Find the most frequently occurring region of widths and heights
%Divide the widths and heights into bins
[bincountsWidths,edgesW] = histcounts(allWidths);
[bincountsHeights,edgesH] = histcounts(allHeights);
%Find the max value in both, then find the value in edges that has the same
%index plus the next one, and this is the desired range of dimensions.
[maxM, edgeMLower] = max(bincountsWidths);
[maxN, edgeNLower] = max(bincountsHeights);
%Check if the maximum value occurs only once
maxOccurrencesWidth = length(find(bincountsWidths(:)== maxM));
maxOccurrencesHeight = length(find(bincountsHeights(:)== maxN));
if maxOccurrencesWidth == 1
    %Occurs once -> take the edges of this bin
    edgesMaxOccurrencesW = edgesW(edgeMLower:edgeMLower+1); %#ok<NASGU>
else
    %Occurs more than once -> find every occurrence of the max value and find the edges of those bins
    a = find(bincountsWidths(:) == maxM);
    edgesMaxOccurrencesW=zeros(length(a),2);
    
    for k=1:length(a)
        %Create an arrray with as many rows as there are length(a)
        %Put the upper and lower edge of each max occurrence in that array
        edgesMaxOccurrencesW(k,:) = [edgesW(a(k)), edgesW(a(k)+1)];   
    end
end
if maxOccurrencesHeight == 1
    edgesMaxOccurrencesH = edgesH(edgeNLower:edgesNLower+1); %#ok<NASGU>
else
    b = find(bincountsHeights(:) == maxN);
    edgesMaxOccurrencesH=zeros(length(b),2);
    
    for k=1:length(b)
        edgesMaxOccurrencesH(k,:) = [edgesH(b(k)), edgesH(b(k)+1)];
    end
end 
% edgesMaxOccurrencesW;
% edgesMaxOccurrencesH;
% --------Remove regions that are not within the "most frequent" range(s)
%Find all widths/heights within the bounds of edgesMaxOccurrences
newAllWidths = zeros(size(allWidths));
newAllHeights = zeros(size(allHeights));
for k = 1:size(edgesMaxOccurencesW, 1)
    for j = 1:size(allWidths,1)
        for l = 1:size(allWidths,2)
            if (allWidths(j,l) > edgesMaxOccurencesW(k,1) && allWidths(j,l) < edgesMaxOccurencesW(k,2));
                newAllWidths(j,l) = allWidths(j,l);
                newAllHeights(j,l) = allHeights(j,l);
            end
        end
    end
end
for k = 1:size(edgesMaxOccurencesH, 1)
    for j = 1:size(allHeights,1)
        for l = 1:size(allHeights,2)
            if (allHeights(j,l) > edgesMaxOccurencesH(k,1) && allHeights(j,l) < edgesMaxOccurencesH(k,2));
                newAllWidths(j,l) = allWidths(j,l);
                newAllHeights(j,l) = allHeights(j,l);
            end
        end
    end
end
%Now I want to find all the regions within the image that do not meet these
%dimension criteria and "delete" them