This thesis presents a video object tracking system based on the active contour model and color information classification. First, the active contour model is applied to detect the target object and the background regions in the initial frame. The pixels of object and the background regions are clustered according to their color information using the k-means clustering algorithm. Then the video objects in the subsequent frames are automatically detected and tracked using temporal differencing and block matching. The moving and stationary regions in a frame are estimated by the temporal differencing. In the moving regions, pixels are obtained their classification from the previous frame using block matching while they are directly received their classification from the previous frame in the stationary regions. Finally, a post-processing procedure is applied to correct the error pixels. Experimental results show that the proposed method provides better performance than the active contour method in video object tracking.
