Authors : Mary Metilda and T. Santhanam
Abstract: The importance of shape as a tool for analysis and classification has led to its study in many diverse fields of science and engineering. In this research, face recognition problem is treated as a 3D shape recognition problem of a free form curved surface. The proposed approach is based on the configural properties of the face that are reflected in the surface structure of the face. Filtering and normalizing techniques are applied to reduce the effect of quantization and also to smooth the curvature. The new-fangled algorithm is introduced to recognize and locate points of local maxima from the smooth curvature and also to reduce dominant points in order to reinforce the efficiency of the face recognizer. ART neural network (Adaptive Resonance Theory) is employed to match the shape of the test image with that of the reference shape stored in the database to discriminate the identity of the input. Experimental results show that the suggested algorithm can extract the shape descriptors of the face with estimable precision and the extorted contour are effectual to enhance face recognition. A reduction of the dominant points is also desirable to augment the competence of the face recognizer, which is comprehended in this algorithm.
Mary Metilda and T. Santhanam , 2007. Facial Recognition Using Curvilinear Shape Descriptors . Asian Journal of Information Technology, 6: 1064-1070.