Structure descriptor for articulated shape analysis

Abstract

Combining the anatomic and volumetric metrics we propose a structure-aware method to effectively detect articulation structures. It helps non-experts to quickly generate structural descriptors which can be used for realistic shape animation and shape understanding. Firstly, the geodesic distance for topological analysis is computed to obtain a topological skeleton. Secondly, we refine the articulation joints with the help of an anatomic and volumetric measurement. The enhanced graph encoded with structural joints provides an affine-invariant and meaningful structure descriptor of articulated shape in a reasonable execution time. A series of experiments have been implemented to show the robustness and efficiency for most articulated shape analysis and shape animation. © 2013 Springer-Verlag.