Novel correspondence-based approach for consistent human skeleton extraction

This paper presents a novel base-points-driven shape correspondence (BSC) approach to extract skeletons of articulated objects from 3D mesh shapes. The skeleton extraction based on BSC approach is more accurate than the traditional direct skeleton extraction methods. Since 3D shapes provide more geometric information, BSC offers the consistent information between the source shape and the target shapes. In this paper, we first extract the skeleton from a template shape such as the source shape automatically. Then, the skeletons of the target shapes of different poses are generated based on the correspondence relationship with source shape. The accuracy of the proposed method is demonstrated by presenting a comprehensive performance evaluation on multiple benchmark datasets. The results of the proposed approach can be applied to various applications such as skeleton-driven animation, shape segmentation and human motion analysis.

Duke Scholars

Author Feng Tian DKU Faculty