Mean-shift video tracking using color-LSN histogram

A texture based object tracking algorithm is presented. The algorithm is an extension to famous mean-shift tracking method. It does not rely on color histogram. It incorporates both color histogram and texture histogram information to model tracking target. In order to extract textural information a novel similarity operator is introduced. It is fast to compute and gray-scale invariant. It would be shown that the proposed method outperforms a similar algorithm. Our experiments shows great improvement over classic mean shift tracker. Also, the algorithm is a quick one in comparison to a similar method.