基于图中心和自动阈值的关键帧提取方法

Key Frame Extraction Based on Graph Center and Adaptive Thresholds

随着因特网的迅猛发展,迫切需要对越来越多的视频信息进行迅速且低成本的管理。而关键帧的使用可以大大减少视频索引的数据量,同时也为查询和检索视频提供了一个组织框架。文章中利用图与视频结构的相似性,将求关键帧的问题转化为求图的P中心问题;同时利用镜头之间的活动度不同,自动选取阈值。实验结果表明提取的关键帧集能够很好地表示镜头的主要内容,且有较低的计算复杂度。

With the Intemet developing, more and more video informations need to manage quickly, Key frame can decrease data quantity in video indexing and provide a framework for video retrieval and indexing ,In this paper, we present a new method of key frame extraction, Using the similarities between graph description and video structure, we can regard the problem of key frame extraction as a problem of finding the P-centre of a graph, At the same time using the difference activities of shots, we obtain an automatic threshold. Experimental results show that the method can produce good representative key frames with low computational cost.