FACE SEGMENTATION, FACIAL FEATURES EXTRACTION AND TRACKING USING SPATIAL FUZZY CLUSTERING METHOD

摘要

The automatic detection of face, extraction of facial features (eyes, eyebrows, mouth and nose) and tracking are crucial to many applications including recognition of human faces and content-based video coding. In this paper we propose to use the spatial fuzzy clustering technique to carry out a two stage process to detect the facial region and other facial features. The facial region is extracted in the first stage of processing based on the membership values of all pixels. The eyes and eyebrows are then located from the extracted facial region. With the geometric face structure, the mouth region and nose are then identified. Experimental results show that the proposed approach works well even for subjects with glasses and beard. For a database size of 134 images, the facial region is correctly segmented for 133 images. A successful rate of 91% for eye location is achieved for subjects without glasses, and 94% for mouth location for subjects without beard.