文档介绍：摘要
随着互联网上多媒体信息的不断丰富,人们对信息的检索需求已不再局限于文本检索和网页检索。人们开始在互联网上检索图片、音频、视频等资料。这些多媒体信息在内容上是非常复杂的,尤其是视频信息。因此,准确地挖掘出视频的内容信息对视频检索是非常重要的。
本文主要研究了基于内容的视频信息检索的相关技术,并在多媒体资料库系统中实现了基于内容的视频检索功能模块。本文研究内容主要包括国内外知名 CBVR 系统,并以这些系统发展趋势作为本文研究方向的依据;分析了视频结构信息的特点,研究了常用的镜头分割算法、关键帧提取算法、镜头聚类算法和基于相关反馈的视频检索技术,并对镜头聚类算法进行了改进;实现了新闻主题字幕检测方法,并针对新闻类视频文件提出一种基于主题、关键帧、视频段的分层反馈视频检索方法。
最后,实现了一个多媒体资料库系统。本平台主要实现了视频文件的存储、视频检索、视频镜头分割、关键帧提取、新闻主题检测等功能来对本文所研究的视频分析处理方法进行评测和验证。实验结果表明,本文改进后的方法是十分有效的。
关键词:视频检索镜头分割关键帧提取新闻主题字幕检测
Abstract
As the multimedia information on the is continuously enriched, people have more retrieval needs for all kinds of information, which are no longer limited to the text retrieval and web retrieval. People begin to retrieve much information on the , such as images, audio and other information. It is very important for the video retrieval to excavate the content of the video information, since the content of the multimedia information is plex.
The content-based video information retrieval technology is studied in this paper, and the content-based video retrieval function modules are realized in the multimedia database system here. First of all, the well-known CBVR system is studied and the development trends of the system is taken as the basis of the research in this paper. Secondly, the characteristics of the video structure information is analyzed, some important technologies, including shot segmentation algorithm, key frame extraction algorithm, shot clustering algorithm and video retrieval technology based on relevance feedback, are researched here. Based on the analysis and study above, the key frame extraction algorithm and the shot clustering algorithm are improved. Finally, a new detection method for the topics subtitles of news, and a hierarchical feedback video retrieval method based on subjects, key frames and video segment, are proposed in this paper.
A multimedia database system platform is designed finally. Many functions, such as the storage of video files, video retriev