incorrect results from bwconncomp

I would like to understand how to correctly use the bwconncomp function, as the results I get do not make sense. Here is my problem. I have a binary image of edges like so:

It has been preprocessed to remove all vertices. I want to identify all the edges individually, so I ran the code:

CC1 = bwconncomp(E1);

It says there are only 13 objects, which looks highly suspect, so I pull up the first object like this:

%create cell array of edge matrices
numPixels1 = length(CC1.PixelIdxList);
edges1 = cell(num_1,1);
for ii = 1:num_1
    pix = CC1.PixelIdxList{ii};
    
    [rows,cols] = ind2sub(size(E1),pix);
    r = max(rows)-min(rows)+1;
    c = max(cols)-min(cols)+1;
    mat = zeros(r,c);
    rows = rows - min(rows)+1;
    cols = cols - min(cols)+1;
    inds = sub2ind([r,c],rows,cols);
    mat(inds) = 1;
    edges1{ii} = mat;
    
end

%show first element
figure;imshow(edges1{1})

And I get this:

I don't understand how counting connected components would actually connect those components before counting, it seems precisely in oppposition to the goal. Please help to clarify the use of this function.

[Matt J]

I get 149 objects. It would be advisable to post E1 in a .mat file, so that we know we are all working on the same image.

[Erin Linebarger]

Sure thing! Just added it

[Erin Linebarger]

aha. just noticed E1 is not a binary image after all, due to the preprocessing to remove vertices, which has resulted in some -1 entries. After correcting this so that E1 is correctly binary, I get 148 objects. Problem solved!

Answer 1

[Image Analyst]

Identify the branchpoints with bwmorph() then mask them out.

bpImage = bwmorph(mask, 'branchpoints');  % Find branchpoints in your binary image called "mask"
mask(bpImage) = false;  % Erase those branchpoints.

[Erin Linebarger]

That worked! Previously I was finding the branchpoints, dilating them to ensure full disconnection, and then subtracting them, but this resulted in some values negative so after I realized this, I had to additionally set all negative values to 0. Your solution is cleaner. (Although I still dilate the mask to ensure all edges disconnected in the 8-connectivity sense.) Thanks!!

[Image Analyst]

Yes, while logical variables CAN be treated as numbers 0 and 1, it's always best to use logical operations to avoid the problem you saw. Subtracting a true (1) from a false (0) will require them to be cast to double instead of logical so it can handle the -1. Then that causes weird side effects like you saw, along with not being able to use the array as a logical index anymore (like I did). So while in some circumstances you can use mathemetical operations, it's always safest to stick with logical operations when dealing with logical variables.

Thanks for accepting the answer.