Abstract: As an effective iterative algorithm, Mean-shift is widely used in object tracking, image smoothing and other computer vision areas. In view of the fact that the standard Mean-shift tracking algorithm is lack of scale adaptation mechanism and Camshift algorithm needs choosing face local sample manually to track accurately before detecting,an adaptive tracking window scale algorithm was proposed for detecting face in video. I}he algorithm adopts illumination compensation and skin color segmentation to rectify the size and location of tacking window. The results of experiments show that the proposed algorithm not only has good real-time capability, but also reduces the location error of the traditional algorithm and achieves more accurate face detection in video.

Key words: Face detection, Object tracking, Scale adaptation, Mean-shift