| With the development of Internet and broadband multimedia business, video becomes one of the mainstream vehicles in modern information dissemination. As video information comes forth largely, how to search the needed information quickly in a mass of data becomes very important. Content-based Video Retrieval is one of the leading methods to get the needed information from the vast amount of video data. The meaning of CBVR is that with the computer which can analyze video contents automatically, users can bring forward requirements according to their own needs.Based on the introduction of the technology of CBVR, the article mainly studies and discusses the following aspects.(1) Key frame extraction. Technology to extract key frame is an important part of content-based video retrieval. In this paper, based on the study of existing key frame extraction algorithm, random algorithm is used in the process of key frame extraction.In the compressed domain environment, firstly divide the two frames to be compared into sub-blocks, and then use randomized algorithm to select a subset of the divided sub-blocks to calculate similarity, and then to select key frames. The method can be guaranteed to meet the inter-frame similarity calculation under a certain precision, significantly reducing the number of sub-blocks needed to compare, thereby enhancing operational efficiency of the algorithm, and experimental results proved to be better.(2) Video abstraction extraction. Video abstraction can reduce the cost of video storage, classification and indexing by analyzing the video content, as well as can improve the efficiency, availability and accessibility of the video. In this paper, information theory is used to calculate the size of mutual information between the adjacent key frames, and then to extract the key frames from the initial key frames to form a video abstraction, and experiment results proved to be better.(3) Design and realization of the system. Based on the above research contents, a retrieval system of CBVR is designed in VS2005, DirectShow and SQLServer2000, and the retrieval based on the browser and sample is realized. |
