A multiresolutional approach for facial motion retargetting using subdivision wavelets

Abstract

We propose a new facial motion retargetting method using wavelet-based multiresolutional analysis of triangular meshes. We define the displacement of the source model as the difference between the neutral and the current expressions of the source model, and decompose it to the displacement of the base mesh and the "displacements" of the wavelet coefficients. We compute the wavelet coefficients of a mesh by using the wavelet basis functions derived from a mesh subdivision scheme. To get the motion of the target model, we add the displacement of the source model to the target model level by level starting from the base mesh. The wavelet-based representation of mesh facilitates the comparison between two expressions of the same model and the comparison between the expressions of the source and target models. © Springer-Verlag Berlin Heidelberg 2005.