Motion Analysis and Reconstruction of Human Joint Regions for Sparse RGBD Images

Abstract

We propose a motion deformation analysis framework that can re-construct the joint region of dynamic human only depending on extremely sparse RGBD images. In the framework, we analyze the deformation mechanism of the joint region from the perspective of its internal material. The non-rigid deformation of a joint region is transformed into the elastic deformation of multiple elements by introducing Finite Element Method. Considering the influence of joint position on the motion deformation analysis, we optimize the joint positions obtained from skeleton tracking prior. Experimental results verify the rationality and effectiveness of our framework.