Skeletonization of Branched Volume by Shape Decomposition

Abstract

We present an algorithm to automatically extract skeletons for branched volumes by shape decomposition. First, a region growing strategy is adopted based on a distance transformation to decompose a volume into several meaningful components with simple topological structures. Then, the skeleton of each component is individually extracted. Finally, the skeletons of all the components are integrated and a structural skeleton of the volume data is obtained, where the structural skeleton is topologically equivalent to the volume. The contributions of the algorithm are: the elimination of the influence of different branches and the accurate skeleton extraction with topological structure of the model due to exact decomposition. Experiments show that this algorithm is applicable to shapes with complex topology.