An improved mean-shift tracking algorithm with spatial-color feature and new similarity measure
Lurong Shen, Xinsheng Huang, Yuzhuang Yan, Shengjian Bai · 2011
The mean-shift algorithm has achieved considerable success in object tracking due to its simplicity and robustness. However, the lack of spatial information often leads to false positives of the color based tracker when the background has a similar color style. Furthermore, the classical similarity measures are not very discriminative. In this paper, an improved mean-shift tracking algorithm with spatial-color feature and a new similarity measure function is proposed. This leads to a very efficient and robust nonparametric spatial-color feature tracking algorithm. The algorithm is tested on image sequences and shown to achieve robust and reliable frame-rate tracking.