摘要
在提出一种新闻视频检索系统结构的基础上 ,介绍了基于标题条检测的新闻场景分段索引方法 ,并阐述了两种新闻摘要的生成策略 .该方法利用新闻节目标题条出现的时空位置线索 ,结合标题关键词的识别 ,建立起一个层次化的新闻视频索引结构 ,并帮助用户通过新闻摘要实现不同要求的视频浏览 .实验证明该方法有较高的检索成功率 ,并且简单快捷 ,为新闻视频检索提供了新的有效途径 .
On the basis of a existing news video retrieval system, this paper introduces a method for news video indexing by scenes and two ways of content based summary, using caption detection. In order to set up a content hierarchy of news video, this method makes use of some particular temporal and spatial characters of captions appearing in the news video. In addition, technology of character recognition can also help the system to give a good annotation according to different news items. This new strategy achieves good results by experiments and it supplies an efficient way for news video retrieval.
出处
《计算机学报》
EI
CSCD
北大核心
2003年第7期859-865,共7页
Chinese Journal of Computers
基金
国家自然科学基金 ( 60 172 0 2 5 )资助
