Segmentation of a head into face, ears, neck and hair for knowledge-based analysis-synthesis coding of videophone sequences

Since in video telephony the image quality in the face is subjectively more important for a human observer than the image quality in other head parts like the hair, the neck and the ears, a knowledge-based analysis-synthesis coder should code different head parts with different qualities. For this, an automatic segmentation of a head into different head parts is necessary. An algorithm for automatic segmentation of a head into face, ears, neck and hair is presented. This segmentation is carried out using estimates for eye and mouth centers, chin and cheek contours, head silhouette and areas with the skin color of the person. The proposed algorithm has been applied to the videophone sequences "Claire" and "Miss America". The segmentation of the heads into head parts is carried out with subjectively high accuracy.