A New Spatiogram Similarity Measure Method and Its Application to Object Tracking

For spatiogram based object tracking, suitable similarity measure is critical. In this paper, a new spatiogram similarity measure is presented. The spatial distribution of the pixels corresponding to each bin is regarded as a Gaussian distribution, where the mean vector and covariance matrix are computed with all pixels belonging to the corresponding bin. Then, the similarity of two spatial distributions is computed with the Jensen-Shannon Divergence (JSD). The similarity of color feature is calculated by using histogram intersection, which is more discriminative than Bhattacharyya coefficient. Both theoretically and experimentally, the proposed measure is stable, and gives superior discriminative power than existing methods, and achieves promising performance in tracking object from single or sequence of images.