基于直方图差分的视频分类方法  

Video Classification Method Based on Histogram Difference

作　　者：孟丽[1] 徐法升[1] 李金屏[1] 

机构地区：[1]济南大学信息科学与工程学院,山东济南250022

出　　处：《济南大学学报（自然科学版）》2007年第2期100-103,共4页Journal of University of Jinan(Science and Technology)

基　　金：山东省高校中青年学术骨干计划(G0204);济南大学科技发展基金(Y0521)

摘　　要：针对当前视频自动分类算法复杂计算量大等问题,提出一种简单的自动视频分类方法,即基于直方图差分的统计方法。获得每个视频的直方图差分曲线后,调整去掉编辑特效造成的虚假峰,得到每个视频的每秒平均镜头切换次数,以此作为广告视频和非广告视频的分类依据,进行C-均值聚类。实验结果表明,该方法以较少的工作量获得了较好的分类结果。Seeing that most of the traditional automatic video classification algorithms are complex and always waste lots of time, we proposed a simple video classification approach of histogram difference statistical method. In this plan, the false peaks caused by gradual shot change in the histogram difference curves have been removed. It is very convenient to find out those advertising videos from others according to the shot change times per second by C - Means clustering method. It has been proved by our practice that this method gains great benefit while saves tremendous time.

关 键 词：自动视频分类 直方图差分 每秒平均镜头切换次数 C-均值聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]