【摘要】 利用无监督聚类算法来提取关键帧是一种常用的方法,但该算法对类别数和初始类划分较敏感,在对视频内容一无所知的情况下,要求预先指定聚类数目是一个很困难的问题·提出一种二次聚类的方法;第1次以镜头内相邻两帧的相似度为数据样本进行聚类(分成两类),计算确定第2次聚类所需的阈值;第2次采用动态聚类的ISODATA算法,以视频序列的帧为数据样本进行聚类,得到最终聚类结果·最后在每类中自动提取距其类中心最近的帧为关键帧·该算法简单且行之有效,无需预定义任何阈值(如聚类数目)·对大量不同特点的视频进行了实验,该算法均取得了较好的实验结果·更多还原

【Abstract】 It is a common method to extract key frames using the unsupervised cluster algor ithm. But the algorithm is sensitive to the initial number of the classes and the initial classification. It is problematic to predefine the absolute number of key frames without knowing the video content. An approach for two times clus te ring is presented. In the first time, the similarity distances of the consecuti ve frames in a shot are clustered into two classes so that the thresholds needed in the second time clustering process can be determined adaptively. In the sec ond time clustering, all the frames in the shot are clustered using dynamic clus ter ISODATA algorithm. Then the frame nearest to the center of its class is aut omatically extracted as one key frame in the shot. It is simple and effective with noneed to predefine any threshold. Experimental results of many videos with di fferent traits demonstrate the good performance of the proposed algorithm.更多还原