Robust object tracking approach using mean shift
一种基于mean shift的鲁棒性目标跟踪方法

Background pixels in object model will induce localization error of object tracking, but in order to let the object contained in object model, it was inevitable to introduce some background pixels in object model. For reducing the localization error of object tracking, a weight was used in object model. The weight indicate the alike degree between background character and object character and could reduce localization error of object tracking introducing by background pixels in object model. The experimental results show the approach has good localization precision of object tracking, and are robust against occlusion.