Robust Distribution Fields Tracking based on Bayesian Mutual Information

Online object tracking is an important issue in machine learning. In this paper, a robust distribution fields tracking algorithm based on bayesian mutual information is proposed. Frist, the similarity based on bayesian mutual information is introduced to measure the distribution fields. The robustness of the bayesian mutual information is enhanced by using the significant area. Then, the target location is determined by integrating the vote maps calculated by matching the feature layers. Finally, an adaptive updating scheme of templates is proposed to overcome the tracking drift problem. The comprehensive evaluations are conducted, and the experiments results show that the robustness and accuracy of our tracker.