摘要
提出了一种电视新闻节目主持人镜头的自动检测方法。该算法可用于新闻节目的自动内容分析和检索。方法分为两步。在第一步中,把屏幕分为两部分:左屏幕和右屏幕,并分别加以检测,然后依据主持人镜头在一个新闻节目中频繁重复出现的特性对候选主持人镜头进行聚类。在第二步中,根据主持人镜头出现的时空特性,利用主颜色直方图检测出正确的主持人镜头。实验结果表明该方法有很高的正确率和适应性。该方法可用于新闻节目内容的自动分析和检索。
A new method is presented to detect anchorperson shots automatically for digital TV news programs.The identification process consists of two steps.In the first step,the screen is divided into two parts:left screen and right screen which are detected respectively,then potential anchorperson shots are grouped by clustering based on the fact that anchorperson shots always present repeatedly within a news program.In the second step,Dominant Color Histogram is used to identify the correct anchorperson shots according to temporal and spatial features of their occurrence law.Experimental results show that the proposed method can detect the anchor shots quickly with high precision.The algorithms are required for automatical parsing and indexing of news program.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第15期183-185,共3页
Computer Engineering and Applications
基金
河南省教育厅基金项目(编号:sp200303099)
关键词
视频信息检索
多媒体数据库
视频内容分析
主持人镜头检测
视频数据库检索
video information retrieval,multimedia database,video content parsing,anchorperson shot detection,video data-base indexing
