Finding number of connected components of a single value of pixel

I have an image in the form of a matrix:

A zoomed-in portion of the upper right part is:

I would like to remove the disconnected black region, i.e. set that particular region's pixel values equal to that of the white region around it. I tried finidng the number of connected components using

CC = bwconncomp(filled_win);

but I am getting only 1 connected component.

CC.Connectivity = 8, CC.ImageSize = [416, 349]
CC.NumObjects = 1, CC.PixelIdxList = 1 X 1 cell

I have attached the relevant .mat file for reference. Can someone please help me with this?

[Ankit Sahay]

The filled_win.mat matrix is not binarized. It can be binarized using

filled_win = imbinarize(filled_win);

Even after doing this,

CC = bwconncomp(filled_win);

doesn't work. Maybe my concept of connected components is wrong, or maybe I am not using the command in the right way. Please help.

Answer 1

[jonas]

If you want to find the two black (false) components, then you need to pass the complement and use 4-connectivity (no diagonal connections).

filled_win = imbinarize(filled_win);
CC = bwconncomp(~filled_win,4);
filled_win(CC.PixelIdxList{2}) = true;

What you found before was the single white component that is the white backgrund and the "boundary", returned as one object (as they are actually connected diagonally).

[Ankit Sahay]

Thanks a lot! This worked perfectly. I am extremely sorry about the late reply.

[jonas]

My pleasure. Dont forget to accept the answer ;)

[Ankit Sahay]

Oh yes..sure!

Answer 2

[Image Analyst]

There are multiple ways to do it. You could just use bwareafilt() to extract the largest blob:

filled_win = bwareafilt(filled_win, 1, 4); % Take largest 4-connected blob.

If the interior blob was the largest blob (not your case) then that could take the interior blob. A different way is to assume that the interior blob has no pixels to the left of the containing blob. Then just take the left most blob, which will have a label of 1

labeledImage = bwlabel(filled_win, 4);
filled_win = ismember(labeledImage, 1); % Take outer blob.

[Image Analyst]

Don't forget to invert your mask! In MATLAB all the functions operate on the white blobs, not the black blobs.