用无监督模糊聚类方法进行视频内容的分层表示

为了在视频数据库中提供有效的视频检索和浏览功能,必须用简明的方式表示视频的内容。由于视频数据具有层次性结构,在镜头边界检测后,可以利用聚类方法按不同的相似性尺度选取代表帧和代表镜头,对视频内容进行抽象概括的表示。文中提出了一种基于无监督模糊聚类对视频内容进行分层表示的算法,它用无监督聚类方法选取镜头的代表帧,并用模糊聚类算法对代表帧进行层次化聚类以选取代表镜头和代表场景。实验结果表明这种方法可以较好地概括视频的内容,方便用户检索和浏览。

Hierarchical Video Content Representation using Unsupervised Fuzzy Clustering Methods

To provide effective video browse,search and retrieval ability,video content should be summarized with compact but proper representation.Because of the hierarchical structure of video data,after shooting boundary detection,cluster method with different similarity metrics can be used to select representative frames and shots and to get an abstract summary of video content.This article proposes a hierarchical video content representation method based on unsupervised fuzzy clustering.It selects representative frames using unsupervised clustering method,then clustering representative frames to select representative shots to obtain a hierarchical representation of video content.Experiments show this method can give a good representation of video content and make video retrieval much easier.