Several research papers have been written on the subject of obtaining facial expression data from images. The feature tracking approach, the pattern tracking approach, and the optical flow approaches are the three major ones. Williams [151] has proposed to acquire facial expressions in real-time using mirrors and dots on the faces. This is perhaps the most detailed performance based facial animation system to date. Terzopoulos and Waters [134] have proposed a method using the ``snakes'' algorithm with facial highlighting. Haibo Li, Pertti Roivainen and Robert Forchheimer [82] have developed a method which extracts affine motions from a facial image and maps it to a FACS-based model for teleconferencing applications. Essa and Pentland [44] use optical flow measurements within a control-theoretic approach to extract facial action parameters from images.