This paper presents an automatic method for simultaneous human face detection and facial feature registration from colur video images. At the first stage, we sue a skin colour Gaussian model to dientify possible face locations under varying paose. Secondly, we compare image patterns with a varying pose face model in terms of shape and texture differences, using a combined feature-texture similarity measure (FTSM). False detections fro mthe first stage are eliminated by setting an appropriate FTSM threshold. Moreover, one can also register the facial features (eyes, nose and mouth) by aligning a prototype face with the unknown pose faces. Experimental results show that hte proposed method can achieve reliable face detection and feature registration under various consitions, including different poses, face appearances, and lighting conditions.
展开▼