Visual tracking via local patches and contextual information

In this paper, a new visual tracking approach via the local patches and the contextual information of the target is presented. In the tracking procedure, the target object is decomposed into a set of patches of equal size and each patch is represented by using intensity and gradient histograms. Then, the likelihood of local patches is defined as the weighted sum of reliability and stability indices, which are applied to evaluate the patches' robustness. Furthermore, the target is represented by using double bounding boxes corresponding to the foreground and background, respectively, which are encoded by HSV color histograms. As this, the drifts can be effectively suppressed by using the contextual information. In the tracking process, the object position is estimated by maximizing the likelihood of the target under the Bayesian framework. The experimental results demonstrate that the proposed approach performs much better than the existing state-of-the-art methods do in terms of efficiency, accuracy and robustness.