Describe a complex shape with simples ones using image analysis

Say I have a complex shape like this

that I want to describe/simplify like this

I was wondering if there was a way using image analysis to do this ? I can find many methods of shape recognition but not any for shape simplification (althoug I'm not an expert and my searching might not be efficient...). I tried using clustering methods (cmdscale on distance between points and linkage/cluster) but the results is not easy to get, especially if I don't know beforehand the number of clusters.

Answer 1

[Image Analyst]

You could threshold the image to produce blobs using imbinarize().

You could separate the blobs using imerode().

You could smooth the blobs shapes using conv2() or imfilter().

[Ouatehaouks]

Thank you for your reply. This is indeed an acceptable answer given my question. In real life though I have lots of various configurations in which I don't really know the size of the link between the secondary blobs from the central one. I guess this would mean a loop with different SE in imerode to see the results and decide, which will probably take some time but I'll give it a try :)

[Image Analyst]

You could try watershed() to separate them.

Watershed transform question from tech support » Watershed transform question from tech support

[Ouatehaouks]

Very promising ! Thank you very much, I'll give it a try (I'll update with the results if it works)