Face Tracking using Multiple Facial Features based on Particle Filter

In this paper, a multiple features face tracking algorithm based on particle filter is proposed. Particle filter can effectively combine multiple face features information which supply robustness in different environments. Meanwhile, our approach makes use of the invariance to rotation and translation of color histogram central moment and statistical characteristic of multiple resolution Sobel Local Binary Pattern (LBP) histogram which shows the local and enhanced global information, then fuses multiple features information by a weight proportion in particle filter framework to propose a new human face tracking algorithm. The experimental results demonstrate the efficiency and effectiveness of the algorithm and present a more robust face tracking performance compared with the method based on single feature.