【作者】 蓝照华；

【导师】 赵进创；

【作者基本信息】 广西大学 ， 计算机应用技术， 2008， 硕士

【摘要】 基于字幕的视频检索是基于内容的视频检索的重要方面,利用新闻字幕进行检索可以大大降低视频检索的复杂度,提高视频检索的速度和准确性,对基于字幕的视频检索技术的研究具有重要的理论意义和应用价值。新闻字幕定位算法是基于字幕的视频检索的核心技术之一,它已成为该领域的研究热点。针对不同的应用背景,有不同的新闻字幕定位算法。本文针对一类具有较复杂背景的图像和视频的字幕定位算法进行了研究。在前人工作的基础上,主要完成以下的工作:(1)对基于字幕的新闻视频检索技术进行了研究,重点研究了新闻视频检索中字幕定位算法及其应用特点。(2)针对一类具有较复杂背景的图像和视频提出了基于C-均值聚类结合边缘检测的新闻字幕定位算法。实验结果表明,该算法具有较好的定位效果。(3)在利用C-均值聚类分割算法进行字幕分割中,针对新闻视频中经常采用的深字白底和白字深底的字幕条,提出固定初始聚类中心的快速C-均值聚类分割算法,以提高基于字幕的新闻视频检索中字幕定位算法的运算速度。(4)探讨了削弱噪声干扰的方法。采用纵向微分运算结合自定义模板(1,1,1,1,1)的中值滤波方法削弱视频图像背景噪声干扰,取得了较满意的实验结果。更多还原

【Abstract】 Caption presents in the video frames plays an important role in understanding the content of a video sequence. Using news caption to do video retrieval can reduce the complexity of video retrieval and increase the veracity of video retrieval, it is very useful for video retrieval based on caption. Caption location algorithm is a core technology of video retrieval based on caption, it has become a hotspot. There are different news caption location algorithms for different application backgrounds.In the paper, caption location algorithm has been researched for a video which has complex background. On the basis of the former, the main works are as follow:(1) News video retrieval based on caption has been researched, focused on the caption location algorithm and its application.(2) An algorithm of news caption location based on C-mean clustering and edge detection is proposed for a video which has complex background. The experimental results show that the algorithm performs well.(3) In order to enhance the velocity of caption location, the rapid C-means clustering algorithm of fixed initial cluster center is proposed for the special caption when doing C-means clustering extraction. (4) Noise reducing technology has been researched. Using vertical differential coefficient and median filter to reduce the influence of complex background, median filter uses the custom template (1,1,1,1,1), it is effective.更多还原