Summary: | 碩士 === 中華大學 === 資訊工程學系碩士班 === 91 === With the improving of living quality, movies play an even more important role in cultures and life. Digital filming data, such as VCDs, DVDs, and those stored in MPEG-4 standards, are emerging. Automatically extracting video summary from these filming data so that users can browse or retrieve movies in a more convenient way is a hot research issue in both the industry and the academic field. To mark their features and to feature the actors within, close-up shots are widely used in traditional film summaries like previews, posters and photos. In this thesis, we propose a new way to recognize close-up shots by means of the C4.5 decision tree classifier. Based on these close-up shots, we can then automatically create new previews, posters and photos. To show the effectiveness of the proposed approach, we also have carried out a series of experiments.
|