Consecutive tracking and segmentation using adaptive mean-shift and graph cut
Junqiu Wang, Yasushi Yagi · 2007
We present an effective tracking and segmentation algorithm in which tracking and segmentation are carried out consecutively. Object tracking in video sequences is difficult since the appearance of an object tends to change. An adaptive tracker that employs color and shape features is adopted to conquer this problem. The target is modeled based on discriminative features selected using foreground/background contrast analysis. Tracking provides overall motion of the target for the segmentation module. Based on the overall motion, we segment object out using the effective graph cut algorithm. Markov Random Fields, which are the foundation of the graph cut algorithm, provide poor prior for specific shape. It is necessary to embed shape priors into the graph cut algorithm to achieve reasonable segmentation results. The object shape obtained by segmentation is used as shape priors to improve segmentation in next frame. We have verified the proposed approach and got positive results on challenging video sequences.