Abstract：Skin color character is widely used for face detection and face tracking. This paper construct a mixed model in both normalized RGB and HSV color space, which can be used to acquire possible skin regions in original image. It designs a sequential flexible morphology operators which can respectively eliminate noises and disturbances, separate face candidates, acquire eye and mouth "hole", and gain contours of faces. The experimental results show that the proposed approach can not only effectively eliminate human hands and arms non-face objects in image, but also smooth away difficulties of face occlusion, pose and orientation, size difference and light variety. Furthermore, this approach is easy to implement, and its computational spending is relatively smaller. So it is suitable to fast face detection.
2004, Vol. 26 
