Automatic 3D Face Model Adaptation with Two Complexity Modes for Visual Communication

Abstract

This chapter introduces a complete framework for automatic adaptation of a 3D face model to a human face for visual communication applications like video conferencing or video telephony. First, facial features in a facial image are estimated. Then, the 3D face model is adapted using the estimated facial features. This framework is scalable with respect to complexity. Two complexity modes, a low complexity and a high complexity mode, are introduced. For the low complexity mode, only eye and mouth features are estimated and the low complexity face model Candide is adapted. For the high complexity mode, a more detailed face model is adapted, using eye and mouth features, eyebrow and nose features, and chin and cheek contours. Experimental results with natural videophone sequences show that with this framework automatic 3D face model adaptation with high accuracy is possible.