extract the tree structure from a noisy data set

Question

I am currently doing a project that attempts to reconstruct the 3D structures from slices, i.e. the tomographic approach. After many steps of image processing, the last task is to connect the “points” (the centroid of the sections) to create the skeletons. The skeletons have been confirmed to be a forest of binary trees.

The problem here is that the data are very noisy. Would you please suggest some techniques to perform data fitting to extract the trees?

Answer 1

I guess a very good resource about that is here (Implementation might be tough, but even reading it will give you a lot of insight about the extraction of 3D centerlines)

Also, please take a look at 1 and 2.