Model Transduction with Mean-value Shape Representation

This paper proposes a novel method, calledmodel transduction, to directly transfer pose between differentmeshes, without the need of building the skeleton configurations for meshes. Different from previous retargetting methods, such as deformation transfer, model transduction does not require a reference source mesh to obtain source deformation, thus effectively avoids unsatisfying results when the source and target have different reference poses.Model transduction is based on two components: model deformation and model correspondence. Specifically, based on mean-value manifold operator, our mesh deformation method produces visually pleasing deformation results under large angle rotations or big-scale translations of handles. We also propose a novel scheme for shape preserving correspondence between manifold meshes. Then, with the above two components, we present the model transduction technique to directly transfer pose between different mesh models. Moreover, we show that the transduction method also can be used for pose correction after various mesh editing operations. Our method fits nicely in a unified framework, where the similar type of operator is applied in all phases. The resulting quadratic formulation can be efficiently minimized by fast solving the sparse linear system. Experimental results show that model transduction can successfully transfer both complex skeletal structures and subtle skin deformations.