基于局部和全局特征相融合的切变镜头自动检测算法  

A SHEAR SHOT AUTOMATIC DETECTION ALGORITHM BASED ON LOCAL AND GLOBAL FEATURE FUSION

作　　者：李秋红[1] 陆蓓[1] 吴海虹[1] 

机构地区：[1]杭州电子科技大学计算机学院,浙江杭州310018

出　　处：《计算机应用与软件》2011年第4期146-149,共4页Computer Applications and Software

基　　金：浙江省科技厅重大专项项目(2010C11049)

摘　　要：提出一种改进的基于局部的亮度差异直方图的方差和全局的灰度直方图帧差相融合的切变镜头自动检测算法,该过程由滑动窗口中相邻两帧的融合特征差值与自适应阈值的比较来自动进行切变检测。提取帧的分块亮度差异直方图的方差作为局部特征,这样既可以避免闪光灯和运动的干扰,又可以增加特征的空间信息;同时提取帧的灰度直方图作为全局特征;接着将二者相互融合,通过和自适应阈值比较来进行切变检测。对大量视频进行实验,结果表明该算法对切变检测具有良好的检测性能,对运动及闪光灯具有较好的鲁棒性,易于实现。An improved shear shot automatic detection algorithm based on local brightness difference histogram variance and global grey histogram difference fused together is proposed.The process automatically detects shear shots according to comparison between the fused feature difference value of two neighbouring frames in sliding windows and the adaptive threshold.Extracting the frame＇s sliced brightness difference histogram variance as local feature not only avoids flashlight and movement disturbance,but also increases feature spatial information;meanwhile the frame grey histogram is extracted as global feature;finally the above two features are fused together to compare with adaptive threshold for shear detection.Experiments with a lot of videos have proved that the algorithm performs well on shear detection,is robust against movements and flashlights and also easy to implement.

关 键 词：镜头边界检测 特征融合 直方图帧差 自适应阈值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]