摘要
新闻视频故事单元关联分析是将新闻视频分割所得的故事单元按照报道的新闻事件所进行的一种特殊聚类。故事单元是新闻视频各层次结构中包含有新闻事件完整语义的结构单元,是对新闻视频进行分析、检索、组织等应用的最佳层次。对新闻视频故事单元关联分析技术进行了综述;按照方法所利用的信息类型将现有方法分成三类:基于文本相似度的方法、基于多模态信息相似度的方法和基于关键帧视觉重复性的方法;讨论了每种方法的特点并对相关技术的发展趋势进行了探讨。
Analyzing relationship of news video stories is a special clustering of stories based on news events. ＂Story＂ is ~ h the unit whic contains full semantic information in hierarchy units. It is the best semantic unit for video analyzing, searching and organizing. This paper gave an overview of existing method and classified them into three categories：the approach based on text similarity, the approach based on multi-modal similarity and the approach based on keyframe visual duplicate. Their advantages and limitation were discussed and compared in detail. At last, this paper presented some trends in this field.
出处
《计算机科学》
CSCD
北大核心
2010年第6期5-10,共6页
Computer Science
基金
863国家高技术研究发展计划(2009AA01Z335)
国家自然科学基金(60802080)资助
