The Research Of Shot-change Detection Based On H.264 Bit Streams

With the development of multimedia technology, digital television technology and the Internet, digital video have been widely used. How to organize and manage the video information effectively and access the information we needed has become an urgent problem. As the important way to get the video information, Video Retrieval is becoming more of a concern. Content-Based Video Retrieval has already become the very focus of researches on the domain of multimedia. It has synthesized all the technologies in the domains of Patterns recognition, Digital Image Processing, Database and Artificial Intelligence, and it has got over the shortcoming of Text-based Video Retrieval. This paper tried to make a deep research into CBVR system and it centered on analysing the key technologies of Shot Boundary Detection.First, we present the background,Research Status and Development of Content-Based Video Retrieval technology, and then sum up the existing shot boundary detection methods in both compressed domain and uncompressed domain, and analyze the advantages and disadvantages of each method. Combined the encoding characteristic of H.264, this paper present a novel algorithm for shot boundary detection for h.264 compressed video based on the existing work. Firstly, extracting intra-frame code partition mode or interframe code prediction mode of the macroblocks to analyze similarities of the neighboring frames, by this way we detect the abrupt shot-changes. We use different weight to different regions of the frame, by this we can prominent core contents of the frame and effectively eliminate the influence caused by the change of the spectator stand of the match or the captions of the films. We widen the gap between the none-transition frames and transition frames by takeing bi-directional-predicted blocks into account. Then, we detect gradual transition detection by annlyzing the average motion intensity and the the number of blocks using intra-coding. It effectively solve the problem of erroneous inspection of global motion or sudden changes in light intensityThe result revealed that such an algorithm required for less arithmetic amount and the effectiveness and accuracy of the method is improved greatly.