| This thesis explores the representation of the human face by features based on the curvature of the face surface. Depth and curvature features have several advantages over more traditional intensity based features. Specifically curvature descriptors (1) have the potential for higher accuracy in describing surface based events, (2) are better suited to describe properties of the face in the area of the cheeks, forehead, and chin, and (3) are viewpoint invariant. Although several researchers have worked on the problem of interpreting range data from curved (although usually highly geometrically structured) surfaces, the main approaches have centered on segmentation of the data into simple surface types based on the sign of mean and Gaussian curvature. These types of methods are not sufficient to classify more natural smoothly curving objects, whose surfaces are not well modeled by piecewise planar, cylindrical, or spherical regions. This thesis details the calculation of principal curvature, the calculation of general surface descriptors based on curvature, and the calculation of face specific descriptors based both on curvature features and a priori knowledge about the structure of the face. These face specific descriptors can be incorporated into many different recognition strategies. We implement two such strategies, one based on depth template comparison and the other based on comparisons in feature space. Both systems show very promising experimental results. |
