A Color-Based Approach to Automatic Human Face Detection
Faculty of Computing, Health and Science
School of Engineering / Centre for Communications Engineering Research
We present a new approach to detecting quasi-frontal faces in color images. Our approach combines a color-based skin detection technique, a face candidate selection scheme using a geometric face model and color-based eye detection, and a face/nonface classification method based on the naive Bayes model. The proposed approach eliminates the computation-intensive step of window-scanning commonly adopted in holistic face detection approaches. With the proposed approach, we also address two important problems in face detection, namely coping with in-plane rotation and detecting faces of different sizes.