3D shape retrieval for articulated models

This paper proposes a novel 3D shape retrieval method for articulated shape models. The proposed method is composed of a topological matching step based on skeleton graph and a geometric matching step based on shape diameter function. In the first step, the curve skeletons of the shape models are extracted and a similarity measure based on skeleton graph is used for 3D shape retrieval. In the second step, a "slim-fat" shape descriptor is calculated and the shape retrieval process can further sort shape models with this shape descriptor. Experimental results show that our method produces an efficient and accurate result.