Model-based and image-based methods for facial image synthesis, analysis and recognition

We review several model-based and image-based methods that we have developed for analyzing, synthesizing, and recognizing facial images. Our model-based methods include a sophisticated, functional model of the human face/head, which incorporates a biomechanical tissue model with embedded muscle actuators, and techniques for applying it to computer animation and expression estimation in video. Our image-based methods include tensor-faces, a nonlinear (multilinear) representation for facial image ensembles that disentangles pose, illumination, and expression effects to improve facial recognition.