新闻视频故事单元分割技术综述

新闻视频的故事单元分割一般采用统计学或者信息论的方法,将新闻节目分割成一系列有各自主题内容的故事单元。这些单元反映的是视频流的高层语义,是建立视频索引的最佳层次。该文对这一技术进行了综述,将现有方法根据利用信息的角度分为3类:单模态的分割方法、多模态融合的分割方法和基于上下文信息的分割方法,并且详细讨论了每一类方法的特点。此外,还分析了一些分割错误的原因和今后的发展趋势。

News Video Story Segmentation Technique: An Overview

As a challenging technique,news video story segmentation divides news programs into a series of stories by statistical techniques or information theory methods.It provides the best semantic units for video indexing.According to the information utilized,this paper gives an overview of the existing methods and classifies them into three categories: the single-modality approach,the multi-modality fusion approach and the context-based approach.Their relative advantages and limitations are discussed in detail and segmentation errors are analyzed.In the end,some trends in this field are introduced.