3D Human Video Retrieval: from Pose to Motion Matching

Abstract

3D video retrieval is a challenging problem lying at the heart of many primary research areas in computer graphics and computer vision applications. In this paper, we present a new 3D human shape matching and motion retrieval framework. Our approach is formulated using Extremal Human Curve (EHC) descriptor extracted from the body surface and a local motion retrieval achieved after motion segmentation. Matching is performed by an efficient method which takes advantage of a compact EHC representation in open curve Shape Space and an elastic distance measure. Moreover, local 3D video retrieval is performed by dynamic time warping (DTW) algorithm in the feature space vectors. Experiments on both synthetic and real 3D human video sequences show that our approach provides an accurate shape similarity in video compared to the best state-of-the-art approaches. Finally, results on motion retrieval are promising and show the potential of this approach.