Particle Filter Optimized by CamShift for Scale Adaptive Face Tracking

In this context, a novel algorithm is proposed for scale adaptive face tracking. Mean-Shift process is applied to Particle Filter (PF) framework, where PF is optimized by Mean-Shift procedure due to its property of accurate tracking and fast convergence. And the accuracy position determination makes window size computation possible in the proposed method. On the other hand, the algorithm maintains multi-hypothesis tracking, so it is still robust. Additionally, an effect mechanism is proposed for checking tracking results, which provides useful information for next frame. In this paper, we have described and evaluated the performance of the proposed algorithm. The experimental results verify its efficiency and robustness for face tracking.