Mean-Shift Tracking Algorithm with Improved Background-Weighted Histogram

The Mean-Shift based visual object tracking has achieved success in the field of computer vision because of its speediness and efficiency. It compute the features of object template and candidate regions by adopting the weighted kernel based color histogram. However, the kernel-based color histogram may not have the ability to locate moving object accurately from the clutter background. In this paper, we propose a robust Mean-Shift object tracking algorithm based on weighted saliency. In order to increase discriminabiltity between object and background preferably and reduce the location error, the saliency of target and background is computed from the histogram bins. By incorporating the weighted saliency into Bhattacharyya similarity metric, an improved weighted background coefficient is defined based on the traditional Mean-Shift. The experiments and the comparison of tracking errors and correct tracking rate show that the effect of tracking is improved.