Registration and Tracking by using Mean Shift Algorithm

Object tracking is the problem of determining (estimating) the positions and other relevant information of moving objects like car, ball or human in an video. There are number of application of tracking surveillance, traffic monitoring, robot vision, animations. Registration is the basic step of tracking. This paper proposes mean shift algorithm which is an iterative method, efficient approach to track a non rigid object. Three different kernels are used for weight distribution, namely Uniform, Gaussian, Epanechnikov kernel. To find the Similarity between the consecutive frames Bhattacharyya coefficient is used. The proposed approach also gives efficient results with different size of searching window