Abstract
Recently, the correlation filters have been successfully applied to visual tracking, but the boundary effect severely restrains their tracking performance. In this paper, to overcome this problem, we propose a correlation tracking framework with implicitly extending search region (TESR) without introducing background noise. The proposed tracking method is a two- stage detection framework. To implicitly extend the search region of the correlation tracking, firstly we add other four search centers except for the original search center in an elegant manner, which is given by the target location in previous frame, so our TESR will totally generate five potential object locations based on these five search centers. Then, an SVM classifier is used to determine the correct target position. We also apply the salient object detection score to regularize the output of the SVM classifier to improve its performance. The experimental results demonstrate that TESR exhibits superior performance in comparison with the state-of-the-art trackers.