Automatic 3D Face Segmentation Based on Facial Feature Extraction

Face can be modeled with a good approximation by a few patches covering some significant regions. Therefore, in many face-related applications, it is critical that face is segmented into patches before being analyzed. This paper presents an automatic 3D face image segmentation algorithm based on feature extraction. Given a 3D face, the 2D texture image is obtained through converting the 3D coordinates to 2D. And then, a robust feature regions extractor is proposed. Both of 2D color intensity and 3D geometric information are utilized simultaneously to locate the facial features. Finally, an automatic 3D face segmentation system is developed based on the extracted feature regions. The experiment result shows that it spends less than 1.5 seconds on segmenting a 3D face. On a database of 500 3D faces the system achieved accuracy more than 95% in locating the major features. The facial patches segmented by our system are regular and could be used for further subdivision and combination with ease.