How to find the skeleton based on medial axis transform 2D

Hi everyone!

I have to measure some length properties in cell images.

I would like to get the "skeleton" of my cells. The output of the bwskel function almost always results in an "overskeletonized" image.

I found out that I can improve this results by previously smoothing the edges of the binary images. However, I feel this process as too subjective (depending on the window size of the smoothing process) and also inaccurate.

On the other hand, the distance transform of the binary image outlines an almost perfect skeleton. Does anyone have an idea of how to obtain the desired skeleton?

I attach an example image of these steps and desired result.

Thanks in advance!

[MARIA RODRIGUEZ SANZ]

Dear John D'Errico,

In response to your comment:

"Not appropriate for MATLAB Answers - Be serious. This is not a question about MATLAB at all. It is purely a question about image processing. So ask it on a forum about image processing. Only once you have some idea of how to solve the problem should you worry about MATLAB implementations."

I work in Matlab, so this question also concerns Matlab and not only image processing. The solution that I'm searching for will requiere Matlab functions, so, it has to do with Matlab.

And, yes, I'm already serious.