基于多特征融合的视频镜头边界检测算法  

Shot Boundary Detection Algorithm Based on Multi-Feature Fusion

作　　者：金凯[1] 丰洪才[2] 张赤[1] 杨亭[1] 

机构地区：[1]武汉工业学院数学与计算机学院,湖北武汉430023 [2]武汉工业学院网络中心,湖北武汉430023

出　　处：《武汉理工大学学报（信息与管理工程版）》2013年第5期673-677,共5页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：湖北省自然科学基金资助项目(2009Chb008;2010CDB06603);湖北省教育厅重点科研基金资助项目(D20101703)

摘　　要：为了建立一种通用、鲁棒的镜头边界检测算法,根据镜头转换的特性和多种视频特征融合的思想,选取了YUV直方图特征、纹理特征和边缘特征进行特征融合并计算帧间差值,采用全局阈值与自适应阈值相结合的方法,对镜头边界检测加以控制,提高阈值选取的准确性。实验结果表明,该算法能够有效地实现视频镜头边界检测,并对闪光、字幕有着较强的鲁棒性。A general and robust shot boundary detection algorithm was established. According to characteristics of lens con- version and the ideal of multiple video features fusion, a shot boundary detection algorithm was proposed based on YUV histo- gram, texture feature and edge orientation histogram. The global threshold was combined with self- adaptive threshold so as to control the process of shot boundary detection and enhance the accuracy of threshold selection. The experiment results show that the algorithm can effectively realize video shot boundary detection and strengthen the robustness of the detection.

关 键 词：镜头边界检测 直方图 特征融合 阈值选取 鲁棒性 

分 类 号：TP371[自动化与计算机技术—计算机系统结构]