Research On Content-based Shot Classification In News Video

With the development of internet and computer technology, more and more multimedia information, especially digital video come to people's live. It is an important task for us to utilize the information effectively. To facilitate people getting the information they want, building a system for video indexing, browsing and search has been the work of researchers.As a basic unit of video, shot can not only be divided into frames but also compose story, which is very important for video analysis. Video shot classification technology bridges the semantic gap between low-level features and high-level concepts, and provides support for the video index and retrieval.News video has a clear structure which makes it play an important role in video content analysis. In this dissertation, news shots are categorized to six semantic types: Commercial, Still Image, Anchorperson, Reporter, Monologue and Others. The"Others"are the rest shots except the first five types in the news video. The types of Commercial, Still Image, Anchorperson and Others are identified respectively with features of themselves. And the reporter and monologue shots are distinguished by conditional random fields (CRFs) model, in which the detection is transformed into sequence labeling problem. The experimental results demonstrate the effectiveness and high performance of the method.