Detection Algorithm Video Retrieval Shot Boundary

With the rapid development of multimedia technology, the multimedia information grew quickly, people need to retrieve their required video from the amounts of video information. How to retrieve the video rapidly and accurately is a focus for the researchers of. The Content-based video retrieval is able to index the video information automatically through video content analysis and processing.The Content-based video retrieval technology is widely used in the area of search engine, digital library, etc.The shot boundary detection is the first step in content-based video retrieval system, the test results will directly affect the retrieval performance of video retrieval system. Shot boundary detection is the process of analysizing the video content, calculating the correlation between video frames and thus classifying the video frames into shots.The shot boundary detection mainly contains three steps:feature extraction, similarity measurement and shot boundary detection.The feature extraction is to descript the video content by exploiting the certain features; the similarity measurement can provide a judgement basis for the boundary detection by calculating the correlation among the video frames and the boundary detection analysize the correlation among the frames to determine the properties of the video frame. The proposed algorithm can improve the detection performance better with increasing the detection time, so that how to tradeoff the detection and detection quality is a focus task.In this paper, the background of the content-based video retrieval system is introduced, then the current existed video retrieval system and the shot boundary algorithm is analysized and the common technology and method of the key three steps in the shot boundary detection is descripted in detail. On this basis, some innovative research work on the shot boundary detection algorithm is proposed in this paper:first, this paper puts forward a shot boundary detection algorithm based on the luminance information and dual complex wavelet characteristic. The algorithm extract luminance information and texture characteristics of the video to descript and analysize the video content more comprehensive thus to improve the detection performance of the shot detection algorithm. Second, Inspired by the compressed sensing theory, a shot boundary detection algorithm based on random projection is proposed which can increase the detection efficiency by using random projection for the image characteristics to achieve dimension reduction. This method reduced the amount of calculation and testing time, while guarantee the algorithm performance.