一种快速新闻视频标题字幕探测与定位方法*

新闻视频字幕包含有丰富的语义信息，尤其是标题字幕，对新闻视频高层语义内容的分析和理解具有 重要作用。利用标题字幕的时空分布特征，提出了一个新闻视频标题字幕的快速探测与定位方法。首先利用标 题字幕持续多帧出现的特点降低所需处理的帧数，然后基于标题字幕的边缘特征和位置特征，标记帧图像的候 选字幕块，对帧序列中的图像进行统计分析，探测出视频中标题字幕的位置及出现消失时间。实验结果表明所 提方法简单有效，能够快速、鲁棒地探测并定位新闻视频中的标题字幕。

Method for fast closed caption detection and localization in news videos

News video caption, especially the closed caption, implies lots of semantics inherently, which is very important for the analysis and comprehension of news videos. This paper proposed a novel method for detecting and locating news closed captions using spatio-temporal distribution feature. Firstly, considering the fact that the closed caption remains in many consecutive frames, only extracted a few representative frames from the frame sequence; then based on edge and location feature of the closed caption,annotated the caption candidate blocks ; finally detect the location and appear-disappear time of closed caption by analysis statistically the image in frame sequence. Experimental results show that the proposed approach can fast and robustly detect closed captions from news video.