摘要： 对海量视频数据库中所蕴涵的语义相关内容进行挖掘分析，是视频摘要生成方法面临的难题。该文提出了一种基于向量空间模型的视频语义相关内容挖掘方法：对新闻视频进行预处理，将视频转化为向量形式的数据集，采用主题关键帧提取算法对视频聚类内容进行挖掘，保留蕴涵场景独特信息的关键帧，去除视频中冗余的内容，这些主题关键帧按原有的时间顺序排列生成视频的摘要。实验结果表明，使用该视频语义相关内容挖掘的算法生成的新闻视频具有良好的压缩率和内容涵盖率。

关键词: 向量空间模型, 主题关键帧, 视频摘要

Abstract: Video summarization is receiving increasing attention to mining semantic contents in huge video databases. This paper proposes a novel emantic content mining approach that mines subject keyframes by an algorithm based on vector space model. After pre-processing, video is transformed into a relational dataset of keyframe classes. Using subject keyframe detection algorithm, it keeps the pertinent keyframes that distinguish one scene from others and remove the visual-content redundancy from video content. The corresponding summary is obtained by assembling them by their original temporal order. Experiments are conducted to evaluate the effectiveness of the proposed approach with summary compression ratio and content coverage. The results demonstrate that meaningful news video summaries is generated.

Key words: Vector space model, Subject keyframe, Video summarization

中图分类号: 

• TP391.4