基于镜头和情节的视频非线性组织  

Shot and Episode Based Nonlinear Organization of Video

作　　者：陆海斌[1] 章毓晋[1] 杨卫平[2] 

机构地区：[1]清华大学电子工程系,北京100084 [2]云南师范大学物理系,昆明650092

出　　处：《计算机学报》2000年第5期548-552,共5页Chinese Journal of Computers

基　　金：国家自然科学基金!( 69672 0 2 9);清华大学"九八五"基金!( EE990 6)

摘　　要：提出了一种新的基于视频单元的视频非线性组织方案 .该方案通过将视频分为情节、镜头、视频流三层 ,以支持视频的高效摘要、存储和浏览 .视频组织采取了由下而上的方法 ,首先通过对镜头单元的检测和镜头关键帧的提取 ,将视频流按镜头进行组织 ;然后在对情节单元划分的基础上 ,根据所提出的镜头代表性准则 ,确定情节的代表帧 ,达到了将镜头按情节组织的目的 .文中着重介绍了所提出的采用分级聚类和模糊分类选取情节代表帧的算法 .根据上述方案对实际视频组织的结果表明 ,该方案可将时序的视频流根据内容有效地重新组合 ,为基于内容的视频浏览和检索创造了条件 .A new nonlinear organization scheme for video is proposed in this paper, which is based on video unit detected from video program.By classifying video program into three layers:episode, shot and video stream, this scheme can support efficient summarization, storage and browsing of video.In this scheme, video organization is made as a bottom up process.First, it is to detect shot unit and extract key frames for organizing video stream into shot.Second, it is to determine representative frames of episode by using proposed criteria in the basis of detected shots.This determination is based on the classification of episode; and it provides a suitable organization of video stream into high level. An algorithm for selecting episode representation frames by using hierarchical clustering and fuzzy clustering is detailed in the paper.The experiment results using the above scheme with real video program show that the proposed scheme can re organize video sequence into an appropriate structure for content based video browsing and retrieval.

关 键 词：视频组织 视频浏览 镜头 情节 多媒体 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]