ABSTRACT
Segmentation of volumetric images by means of 3D deformable meshes is a widespread technique. Extracting the skeleton of such meshes can present great interest for modeling the shape of the segmented area. In this paper, we propose a new method performing fast approximation of skeleton points of a triangular mesh, driven by greedy energy-minimization technique. The skeletonization method uses available data relative to the mesh, such as geometrical and topological structures as well as motion information.
