摘要： 提出用于新闻视频主持人镜头检测的基于扩展人脸区域的模板匹配方法。利用一段新闻节目中主持人服装不变的重要线索，在线提取主持人镜头的扩展人脸区域模板，采用分块HSV颜色直方图作为模板参数，利用该模板对新闻视频镜头中检测出的扩展人脸区域进行匹配，根据匹配结果确定主持人镜头。实验结果表明，该方法计算简单、检测精度高、实时性强，具有很好的通用性。

关键词: 主持人镜头检测, 扩展人脸区域, 信息融合

Abstract: This paper proposes a new method to detect anchorperson shots automatically in news video based on extended face region. In a single news video, the anchor wears the same dress through out the entire program. It takes this fact as the cue, which is independent of the image background as well as the location of the anchor. The anchor extended face regions are extracted as templates on-line and the grid HSV histogram is taken as the template parameter. The well-defined extended face regions are used to match the potential extended face regions from the shots to identify the anchorperson shots. Experimental results show that the new method is efficient, and the detection accuracy is improved compared with the classic methods.

Key words: anchorperson shots detection, extended face region, information fusion

中图分类号: 

• TP391