采用多特征融合的镜头边界检测方法  

Approach of shot-boundary detection based on multi-feature fusion

作　　者：刘群[1] 江伟[1] 吴渝[1] 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065

出　　处：《计算机工程与应用》2010年第13期171-174,共4页Computer Engineering and Applications

基　　金：重庆市自然科学基金No.2007BB2386;重庆市教委应用基础研究项目(No.KJ070504)~~

摘　　要：提出了一种特征融合的镜头边界检测方法。HSV色彩直方图进行镜头边界检测是一种常用、有效的方法,该文融合了视频帧的HSV直方图特征、边缘特征和纹理特征,计算出不连续帧。最后,采用Kohonen网络自组织网络对不连续帧值进行聚类得到镜头边界。实验结果表明该方法不仅是可行和有效的,也解决了需要阈值的问题。An approach of shot-boundary detection by fusing feature is proposed.HSV histogram is a common and effective method for shot boundary detection.In this paper,discontinuous frame is obtained based on HSV histogram,edge orientation histograms and texture feature firstly.Then the Kohonen self-organized network is used to receive the shot boundary by clustering the discontinuous frame.The results of experiments show that this proposed approach can detect shot effectively and doesn＇t need threshold.

关 键 词：特征融合 镜头边界检测 自组织网络 色彩直方图 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]