Three-dimensional modeling of a non-rigid object employing a single camera

This paper proposes a technique for recovering 3-D shape of a non-rigid object employing a single camera. Normally two or more cameras are necessary for recovering 3-D shape of a non-rigid object. Shape recovery of a non-rigid object employing a single camera has recently been studied by a few researchers, but they deal only with small deformation. The technique proposed in this paper achieves the shape recovery of larger deformation by regarding a non-rigid object as a set of rigid parts. Each rigid part recovers its 3-D shape by the application of the factorization method and they are connected mutually to represent an overall deformation of the object. The present technique also considers the shape recovery of an occluded part caused by ego motion. Effectiveness of the proposed technique is shown by experiment.