Prior knowledge of the canonical structure of the human face can aid in various automated faceprocessing tasks. In this project we have developed a statistical appearance model for faces and are exploring its application to several problems: stylized face rendering, caricature, and reconstruction of occluded face images.