| With the development of science and technology, a large number of handheld camera electronic equipments using in human life. As long as people enjoy abundant multimedia information, they also bear so many troubles. Video information grows so quickly so that the technology’s development can’t keep up with the demand such as storage transfer, inquires etc. Due to it’s vivid and intuitive features, video data stands an important position in multimedia information, Effective manage and retrieve video information has become a focus of research.The text-based video retrieval exhausts a great deal of manpower, and the video descriptions are very subjective, therefore the content-based video retrieval (CBVR) has become a hot topic in recent years. CBVR is to extract the features that can represent the video content and to retrieve interested video streams from the enormous video database through matching of patterns based on content and context of the video. CBVR key technology contains shot division, the extraction of key frames, evideo content feature extraction, video organizational structure, the retrieval results feedback mechanism and so on. Two key techniques include shot segmentation and key frame extraction have been studied in this thesis.On the shot segmentation, this thesis puts forward the algorithm which is based on the color space and adaptive threshold. The difference between adjoining frames calculated by blocked and weighted statistical use YUV color space. Two adaptive threshold values are used to judge gradual change and abrupt change. The ratio of interval frames is used to eliminate some misdetection brought by noise. A continuous window value method is used to detect the gradual change.On the key frame extraction, the thesis puts forward a kind of image entropy and the edge matching rate key frame extraction method. First, the shot will be divided into sub-shots by the mutual information, then use image entropy to find out frames which contains the most abundant content in the sub-shots sets,and will get a candidate key frame set.And then the edge matching rate will be used to eliminate a few redundant frames to get the key frame set.Finally, we summarized the content of the research, and point out some deficiencies and the tasks of the next phase. |
