Face tracking using color histograms and moment invariants

This paper presents a face-tracking algorithm based on particle filter framework. Firstly, a target state is obtained using conventional color histograms. Secondly, a tracking method is proposed on the basis of seven moment invariants, and another state vector is also computed using this approach. Furthermore, the weights of the two state vectors are computed according to Euclidean distance between the predictive state and the two states obtained in previous steps. Finally, the weighted average of the state vectors gives the exact state of the target. Experimental results over a set of real-world sequences show that the proposed method outperforms single-feature solution. In particular, the weakness of losing target is overcome, which improves the reliability and robustness of the tracker.