ABSTRACT
A varied but limited set of experiments were performed to test our analysis-by-synthesis technique in estimat ing different facial movements such as pose movements, eye movements, mouth movements, as well as head translations, rotations, and scales. The novel sequences involved changes in lighting, position, scale, rotation, background, clothing, and hairstyle. It should be noted that to change his hairstyle, the author shaved his head! In each illustration, a few frames from the novel sequence are juxtaposed against a few frames from the synthesized sequence. The analysis parameters ex tracted by the algorithm are also shown.
