A Combined Skeleton Model

Skeleton representations are a fundamental way of representing a variety of solid models. They are particularly important for representing certain biological models and are often key to visualizing such data. Several methods exist for extracting skeletal models from 3D data sets. Unfortunately, there is usually not a single correct definition for what makes a good skeleton, and different methods will produce different skeletal models from a given input. Furthermore, for many scanned data sets, there also is inherent noise and loss of data in the scanning process that can reduce ability to identify a skeleton.

In this document, I propose a method for combining multiple algorithms' skeleton results into a single composite skeletal model. This model leverages various aspects of the geometric and topological information contained in the different input skeletal models to form a single result that may limit the error introduced by particular inputs by means of a confidence function. Using such an uncertainty based model, one can better understand, refine, and de-noise/simplify the skeletal structure. The following pages describe methods for forming this composite model and also examples of applying it to some real-world data sets.