Finding Connected Components of Superpixels?

If I know the idx of the following figure to be [ 261 325 331 333 339]. What sort of algorithm can I use that would detect that idxs [325 331 333 339] are all connected to one another, and that 261 has no connected components? The end goal being detecting regions that have 3 or more connected superpixels.

One method I am thinking of is using bwconncomp, which will return the 2 connected objects. And then for each connected object, check weather the idxs are 'contained' in it. I think this might not be a good solution though. Please any feedback is welcome. Thank you all kindly.

[Image Analyst]

Why are you computing superpixels in the first place? What would you do if you knew the adjacent blobs? To help me understand what you really want to do, please attach your original gray scale image and tell us what you want to find or measure in it, or how you want to change it's appearance.

[Suki Sandhu]

Its for a university class of mine where I am to follow the algorithm presented in this paper: 'Segmentation of Overlapping Cervical Cells in Microscopic Images with Superpixel Partitioning and Cell-wise Contour Refinement'.

Specifically, as mentioned in the paper:

First, clustered superpixels as shown in Fig. 5 (a), are reduced by rejecting the brightest superpixel iteratively. Since the nucleus often contains at most two superpixels, a superpixel cluster containing more than three superpixels can be considered to have outlier candidates in it. Superpixel clusters are extracted by connected component labeling and for each cluster containing more than three superpixels, a superpixel with the highest mean intensity is rejected. The cluster extraction and superpixel rejection are iteratively repeated until no clusters are extracted. (Hansang Lee ; Junmo Kim)

Attached is a synethic image I edited, such that it does create nuclei that are segemented into more than 3 superpixels. In this case, I would want the algorithm to remove 2 superpixels attached to the round nucleus at the top. I hope this makes sense. Thank you.

Answer 1

[Image Analyst]

There is a superpixels() function in the Image Processing Toolbox. Using that could significantly simplify your homwwork assignment.