Human Body Motion Capture via 3D Graph-cuts

摘要

In this paper, we propose a new algorithm for markerless vision-based human body motion capture. In our method, motion capture or pose estimation is considered as an energy minimization problem, and the energy Junction construction is converted to a 3D graph construction. During this procedure, a human model with a certain pose will be taken into account. At last Powell algorithm will be used to optimize the result of the energy function.