Abstract: In order to build up a fast and effective system for face detection by combining the skin color information with facial features, the paper proposes a Bayesian detection algorithm based on the skin model. There are mainly two steps in our algorithm: skin color detection and facial feature detection. The former models skin regions to the Gaussian Mixture Model and sets up rules for skin detection. Meantime, a new method based on the cluster validity function is proposed to determine the optimal numbers of GMM components so as to improve the precision of detection. To speed up the facial feature detection, a diamond search algorithm is introduced in the Bayesian classifier. Experimental results illustrate the high detection precision and low false alarm of the proposed algorithm. Moreover, it can realize the face detection in real time basically.

Key words: skin segmentation, GMM, MPFD, Bayesian, diamond search