Comparative analysis of mean shift in object tracking

Object tracking is a exploring field having a broad collection of applications in actual world. The major steps for tracking a target object include detection of interested moving objects and tracking of such objects from frame to frame in a video. Mean shift based object tracking has received much importance because of its assets such as real time, robust and easy to implement. Advanced versions of this algorithm are developed which differ in template updating for representing target objects in different video sequence. This paper performs comparative analysis of basic mean shift algorithm with two enhanced versions of mean shift i.e., BWH and CBWH. The experimental results on bench mark databases show that CBWH target representation leads to faster convergence and more proper localization than other target representations.