Human face tracking with adaptive facial orientation template by mean shift algorithm

Mean shift algorithm has been used in human face tracking during past years and exhibited robust performance compared with other methods. However, the existed mean shift methods usually track faces with vertical orientation and the tracking effect is deteriorated when the facial orientation changed. In addition, it often adopts the color histogram in RGB color space, which is sensitive to lighting variations. In this paper, we present an adaptive facial orientation template for face tracking in YCbCr color space and the experimental results show that the tracking is more efficient to adapt the facial orientation and lighting variations than current methods