Abstarct: Object tracking is one of important fields in machine vision. The aim of object tracking is positioning the object or objects in frxames of a video sequence. In some applications extracting other information of moving object such as the region of the object is also needed. In this thesis we address two main problems in an object tracking system namely object representation and process of tracking. One important topic in an object tracking system is way to represent a moving object. In this thesis a new method for representing object in mean shift tracker is proposed. The existing methods use the statistical color information to represent an object. In the proposed method we use image gradient information beside the color information to improve the performance of existing object trackers. We have shown in presence of similar colors in object image and background, color baxsed tracker fails to track the interest object while the proposed method performs successfully. In case that the object movement is very fast particle filter baxsed tracker performs superior to mean shift algorithm. In the second part of the thesis we propose an object tracking algorithm baxsed on particle filtering in which particles are partially moved baxsed on mean shift algorithm. The result of experiment on tracking samples indicates the proposed method runs quicker and more accurate than trackers baxsed on conventional particle filtering. Moreover comparison between the proposed method and the mean shift tracker demonstrates that when the interest object moves very fast the proposed tracker successfully tracks the object while the mean shift tracker fails. Whereas it keeps to some extent the desirable quality of mean shift tracker named high rate of execution.