Adaptive bandwidth object tracking using sparse approximation

In this paper, a novel adaptive bandwidth object tracking using sparse approximation is proposed in mean shift framework. In this framework, occlusion, illumination and other challenging issues are addressed seamlessly in real time through bandwidth approximation. To find the tracking target at a new frame, a two step method is used iteratively to find the most probable target candidate position. The first step is to find an appropriate bandwidth which best describes the region range of the object using sparse representation, and the second step is to find the target position using mean shift object tracking framework which has been given the appropriate bandwidth in the first step. In addition, we improve the convergence properties of the mean-shift by approximating the region range of target in the iterative update mechanism. The proposed approach shows excellent performance in comparison with originally proposed trackers.