Abstract
In this paper, key components for 3D video retrieval are presented. A shape feature vector is generated from each frame of 3D video by our modified shape distribution algorithm. Based on the feature vectors, motion segmentation and similar motion retrieval can be achieved efficiently.