Fast face tracking using parallel particle filter algorithm

This paper proposed a multi-cue based face tracking algorithm with the help of parallel multi-core processing. Due to illumination and occlusion problems, face tracking usually does not work stably based on a single cue. Three different visual cues, color histogram, edge orientation histogram and wavelet feature, are integrated under the framework of particle filter to improve the tracking performance considerably. To handle the huge amount of computation cost resulted from the introduced multi-cue strategy, a map-reduce thread model is designed to parallel and speed up the observation steps. Besides, an online updating strategy makes our algorithm adaptable to some slight face rotations. The experimental results demonstrate that our proposed face tracking algorithm works robustly for cluttered backgrounds and different illuminations. The multi-core parallel scheme achieves a good linear speedup compared to the corresponding sequential algorithms.