Significant Face Detection Algorithm In Video

Automatic face detection and tracking is an important issue in the field of computer vision. As one of the most intuitionistic and typical biometric features, face is a representative clue, whose presence and authentication provide semantic information to video indexing and relative technology.This paper systematically and deeply studies the automatic face detection and tracking in color image sequences. A Bayesian face detection algorithm based on skin model is presented. In this algorithm, the color information and the structure information of human face is considered to full extend, and the complicated face detection system is simplified to two main steps, skin detection and facial feature detection. The former models skin regions with Gaussian mixture model (GMM) and sets up rules for skin detection. Meantime, a new method based on cluster validity function is proposed to determine the optimal numbers of GMM components so as to improve the precision of detection. In the other hand, to speed up the facial feature detection, a diamond search algorithm is introduced in Bayesian classifier. The experiment results illustrate the high detection precision and low false alarm of the proposed algorithm.In addition, a face tracking method based on eyes location is presented by extending the face detection from image into video sequence. The algorithm can track the significant face defined in news video efficiently, which provides solid foundation for video processing in high-level concept.