Use of a Novel Concept of Potential Pixel Energy for Object Tracking

In this paper, we propose a new method for kernel-based object tracking. Definition of union image blob and mapping it to a new representation which we named as potential pixels matrix are the main part of tracking algorithm. The union image blob is constructed by expanding the previous object region based on the histogram feature. The potential pixels matrix of union blob is used to obtain an algebraic equation for tracking the location of the kernel. We demonstrate that tracking accuracy is independent of the previous object region and it takes effect only from the expanded area of the union blob. To eliminate the background information, we propose a new method which is performed in two stages. At first, the effect of background of expanding part is reduced using a threshold distance. Then, the expanding part is divided into two parts. Tracking equations and a similarity criterion are used for each part to detect the background and target regions. We demonstrate thatthe background removing of proposed method has better performance than the mean shift tracking. Also, for better cancellation of background, the segmentation of object is used. We demonstrate the capability of the proposed method for several image sequences.