基于BAMH的视频镜头边界检测  

BAMH-Based Shot-Boundary Detection

作　　者：祝轩[1] 葛宝[1] 彭进业[1] 李岩[1] 

机构地区：[1]西北大学电子科学系,陕西西安710069

出　　处：《小型微型计算机系统》2004年第9期1652-1655,共4页Journal of Chinese Computer Systems

基　　金：陕西省科技研究发展计划 ( 98k0 7-G4)资助 ;西安市科技攻关计划项目 ( GG990 7)资助

摘　　要：镜头边界检测是进行视频内容分析的首要步骤 .提出一种新的基于分块向量角和模值二维直方图 (BAMH)的镜头边界检测方法 ,将视频帧分割为多块 ,分块统计向量角和模值二维直方图作为计算帧间不连续值的特征 ,采用自适应阈值分割 ,检测镜头边界 .实验结果表明 ,本文方法的检测性能优于全局向量角 -模值直方图方法 ,更优于Dividing a video sequence into shots was the first step towards video content analysis and content based video browsing and retrieval. A novel method for shot boundary detection was presented based on block angle module histogram (BAMH). There, each video frame was divided into several blocks, angle module histograms are computed and normalized for blocks, followed by computing the distance between histograms of consecutive frames. Then an adaptive threshold was selected for detecting shot boundary. Experiments show that the performance of our method was higher than that of the HSV space histogram or globe angle module histogram.

关 键 词：镜头边界检测 分块向量角模值直方图 颜色直方图 基于内容的视频检索 反对称双正交小波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]