基于颜色内容一致性的视频篡改检测  

Digital Video Forgery Detection Based on Color-content Congruency

作　　者：李海涛[1] 林新棋[1,2] 吴鹏[1] 高智[1] 詹晓珊 

机构地区：[1]福建师范大学数学与计算机科学学院,福建福州350117 [2]福建师范大学网络安全与密码技术重点实验室,福建福州350117

出　　处：《福建师范大学学报（自然科学版）》2014年第2期26-33,共8页Journal of Fujian Normal University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61070062;11071041);福建省教育厅资助项目(JA12075;JA10064;JB11036);福建省科技厅高校产学合作科技重大项目(2012H6006);福建省高等学校科技创新团队(IRTSTFJ;N.J1917);福建师范大学网络与信息安全关键理论和创新团队(IRTL1207)

摘　　要：研究假设视频同源帧插入、异源帧插入、替换、删除篡改能够导致视觉内容发生变化,并致力于以底层特征为基础建立这种微小变化数学模型.首先,提取RGB 3个通道的彩色直方图作为底层特征,采用直方图相交法定义视频帧之间的颜色内容一致性值;其次,采用拉依达准则设定阈值,一旦待检测视频帧的颜色内容一致性值小于自适应阈值,则判定该帧被篡改了.实验结果表明该假设是合理,有效的,所设计的算法能够检测出道路监控视频、背景相对复杂视频中的同源帧插入、异源帧插入、替换、删除的篡改.Research supposes that the forgery of digital video, such as similowsourced or di versesourced frame insertion, replacement, deletion, can lead to a tiny change about visual con tent. The aim is devoting itself to establish the changing model based on the lowlevel features. First, the three channel histograms of RGBcolor space are extracted as the lowlevel features, and the histogram intersection is using to compute the value of colorcontent congruency between video frames. Secondly, the adaptive threshold is determined by Pauta criterion. If the value of colorcon tent congruency about undetected video frame is less than the adaptive threshold, the tamper is founded. The experimental results show that this assumption is reasonable, and effective, and the algorithm can successfully detect the some similaror diversesourced frame insertion, replacement, deletion tampers under the road monitoring video, or the video of relative complex background.

关 键 词：视频篡改 彩色直方图 一致性值 自适应阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]