Facial Cloning for Online Interactive Systems

摘要

In this work, we address the problem of robust real-time facial cloning for online interactive systems. The main contributions of the paper are as follows. For face analysis, we develop an extended 2D+3D Active Appearance Models (AAM) based 3D face and expression tracking framework. Moreover, we introduce a robust shape initialization method to handle agile head motion and a penalty function to avoid unallowable mouth shapes. For face synthesis, we generate online real-time 3D expression by blending the expression bases, and these expression bases are created through an offline expression mapping algorithm.