Extraction And Tracking Of Moving Video Images

Video object detection and tracking technique has been in many fields in the past decades. It is a hot problem in this domain and has been increasingly attracting much more attention. In this thesis, an algorithm for automatic detecting and tracking video object in color motion images is introduced. The algorithm makes use of the merits of color-segmentation and motion-one respectively. It adopts data fusion technique for multi-frame information and multi-features and can automatically detect motion object in video images with still background. The object detected has accurate contours. This method can be used in coding video system, videophone,TV conference and tele-education.Then, after the object tracking algorithms based on Kalman filter is studied and the tracking errors produced by the traditional methods are analyzed, a tracking error modified method is presented. As a result, the status-break-effect is lessened in some extent. The experiments show that this method can trace the motion object quickly.Finally, based on the algorithms presented above and combined with the complexional features, automatic human-face detection and tracking system is established. The simulation results show that the system has the ability of detecting and tracking a human face in a motion images.