计算机应用

• 图形图像处理 • 上一篇    下一篇

融合多特征的新闻故事自动分割

梁学战 朱明   

  1. 中国科学技术大学自动化系 中国科学技术大学
  • 收稿日期:2008-11-03 修回日期:1900-01-01 发布日期:2009-04-01 出版日期:2009-04-01
  • 通讯作者: 梁学战

News story automatic segmentation based on multi-feature fusion

Xue-zhan LIANG Ming ZHU   

  • Received:2008-11-03 Revised:1900-01-01 Online:2009-04-01 Published:2009-04-01
  • Contact: Xue-zhan LIANG

摘要: 新闻视频是由一系列的新闻故事构成的,准确地对新闻故事进行探测与分割将对新闻视频的自动检索与语义的理解产生重要作用。通过对新闻视频的结构特征进行分析,提出了融合静音、镜头切变、主持人特征和文本信息等多种特征的新闻故事探测与分割的方法。通过对不同的新闻视频进行实验,获得了平均95.2%的探测准确率。实验证明,提出的方法能够较好地解决新闻故事分割的任务。

关键词: 新闻视频, 结构特征, 故事单元, 主持人

Abstract: News video is composed of a series of news items. It is very important for content-based analysis, indexing retrieval of news video to detect and segment the news items accurately. Analyzing the structure feature of news video, this paper detected and segmented the story items in the news video by using multi-feature such as silent points, shot boundary, anchorperson and theme caption. Also the study used different news videos in our experiment. The experimental results show that the proposed algorithm has high detection accuracy rate, and it can accomplish the task of segmenting news stories.

Key words: news video, structure feature, story unit, anchorperson

中图分类号: