B05: 3D Mesh Simplification for Deformable Human Body Mesh Using Deformation Saliency

Tianhao ZHAO, King Ngi NGAN, Songnan LI

Abstract:
3D mesh of human body is the foundation of many hot research topics, such as 3D body pose tracking. In this topic, the deformation of the human body mesh has to be taken into account because of various poses of the human body. Considering the time cost of the body deformation, however, it"s impractical to adopt a high resolution body mesh generated from scanning systems for the real-time tracking. Mesh simplification is a solution to reduce the size of body meshes and accelerate the deformation process. In this paper, we propose a mesh simplification algorithm using deformation saliency for such deformable human body meshes. This algorithm is based on quadric edge contraction. The deformation saliency is computed from a set of meshes with various poses. With this saliency, our algorithm can simplify the 3D mesh non-uniformly. Experiment shows that using our algorithm can improve the accuracy of body pose simulation in the simplified resolution compared to using classical quadric edge contraction methods.