Perceived Visual Motion Descriptors from MPEG-2 for Content-Based HDTV Annotation and Retrieval

Efficient Content-based retrieval of image and video databases has become an important industrial and consumer application due to rapid and unprecedented proliferation of compressed image and digital video data on the Internet and corporate intranets, and due to the launch of high defintion television (HDTV) broadcast in 1998. We have developed the first HDTV video content management system to date that automatically analyzes motion occurring in MPEG-2 encoded videos within the compressed domain itselfm abd oridyces vudei descruotirs characterizing the global visual motion for content-based video labels can be directly incorporated as annotation indexes into a video database or be used to construct higher level event descriptions of videos. Results from our ongoing experiments with tens of thousands of frames obtained from several MPEG-1,2 video streams of various genres demonstrate the good performance of our system in terms of motion identification accuracy and computational efficiency.