A real-time multi-view face tracking in videos

In this paper, we propose a novel and fast multi-view face detection algorithm for detecting face in color video sequences. This algorithm can be integrated into a real-time surveillance or a video retrieval or indexing system. The skin-color regions are found rapidly by utilizing the advantages that skin color is not sensitive to view and the algorithm has a small amount of calculation. As the face appears elliptical shape, the main axis´ direction of candidate face region is calculated by K-L transform. Finally, the candidate face region corrected by rotating is scanned by trained Adaboost classifier as to achieve the detection of color faces. The experiments show that the proposed algorithm runs fast with good detection accuracy.