如何修改MATLAB连通域标记代码，清除超尺寸亮像素簇？

Got it, let's break down how to modify your code to remove over-sized bright pixel clusters. Your current code already does solid 8-connected component labeling—we just need to add a few steps to count cluster sizes and zero out the ones that exceed your threshold.

Key Modification Steps #

After your existing labeling loop finishes assigning IDs to all clusters, we'll:

1. Count how many pixels belong to each unique cluster ID
2. Identify which clusters are larger than your specified size (e.g., 12 pixels)
3. Set all pixels in those large clusters back to 0 in your output matrix B

Modified Full Code #

visited = false(size(binary_image)); % Track which pixels have been processed
[rows, cols] = size(binary_image);
B = zeros(rows, cols); % Output matrix to store cluster IDs
ID_counter = 1; % Keep track of unique cluster IDs

% Original connected component labeling loop
for row = 1:rows
    for col = 1:cols
        if binary_image(row, col) == 0
            visited(row, col) = true; % Mark dark pixels as visited immediately
        elseif visited(row, col)
            continue; % Skip pixels we've already processed
        else
            stack = [row col]; % Initialize stack with the current bright pixel
            while ~isempty(stack)
                loc = stack(1,:);
                stack(1,:) = []; % Remove the top element from the stack
                if visited(loc(1),loc(2))
                    continue;
                end
                visited(loc(1),loc(2)) = true;
                B(loc(1),loc(2)) = ID_counter; % Assign current ID to this pixel
                % Check all 8 neighboring pixels
                [locs_y, locs_x] = meshgrid(loc(2)-1:loc(2)+1, loc(1)-1:loc(1)+1);
                locs_y = locs_y(:);
                locs_x = locs_x(:);
                % Filter out pixels that are outside the image bounds
                out_of_bounds = locs_x < 1 | locs_x > rows | locs_y < 1 | locs_y > cols;
                locs_y(out_of_bounds) = [];
                locs_x(out_of_bounds) = [];
                % Filter out already visited pixels
                is_visited = visited(sub2ind([rows cols], locs_x, locs_y));
                locs_y(is_visited) = [];
                locs_x(is_visited) = [];
                % Filter out dark pixels (0s)
                is_1 = binary_image(sub2ind([rows cols], locs_x, locs_y));
                locs_y(~is_1) = [];
                locs_x(~is_1) = [];
                % Add valid bright neighbors to the stack
                stack = [stack; [locs_x locs_y]];
            end
            ID_counter = ID_counter + 1; % Increment ID for the next cluster
        end
    end
end

% -------------------- New code to filter large clusters --------------------
max_cluster_size = 12; % Set your desired size threshold here

% Count the number of pixels in each cluster
cluster_sizes = accumarray(B(:), 1);
% cluster_sizes(n) will hold the pixel count for cluster ID = n

% Find all cluster IDs that exceed the size threshold
large_cluster_ids = find(cluster_sizes > max_cluster_size);

% Zero out all pixels belonging to over-sized clusters
for id = large_cluster_ids
    B(B == id) = 0;
end

Optional: Re-label Clusters for Continuous IDs #

If you want the remaining clusters to have consecutive IDs (no gaps from removed clusters), you can add this final step:

% Reassign IDs to be sequential (e.g., IDs 1,3 become 1,2)
[B, ~] = labelmatrix(labelmatrix(B));

This uses MATLAB's built-in labelmatrix to clean up the ID numbering.

Quick Explanation #

• accumarray is a efficient way to count pixel counts per ID: it groups all pixels by their label and sums the 1s we pass in (each 1 represents one pixel).
• The loop over large_cluster_ids directly targets pixels in over-sized clusters and sets them back to 0, which exactly matches your requirement of converting those clusters to dark pixels.

内容的提问来源于stack exchange，提问作者user11459362