Techniques for Realistic Facial Modeling and Animation

This paper describes a methodology for constructing and animating realistic models of human faces. We present three physically-based techniques. The first, an adaptive meshing method, is able to create nonuniform facial meshes from high resolution data acquired by scanning a subject with a laser range sensor. Starting with a nonuniform mesh, the second method constructs a realistic model of the subject’s face and head. The face model includes synthetic facial tissue and muscle actuators based on anatomical and biomechanical considerations. From video sequences of a subject performing expressive articulations, the third method estimates facial muscle contractions and inputs these as dynamic control parameters to the model in order to yield realistic, performance-controlled facial animation.