利用内容连续性的数字视频篡改检测  

Digital video forgeries detection based on content continuity

作　　者：黄添强[1,2] 陈智文[1] 苏立超[1] 郑之[1] 袁秀娟[1] 

机构地区：[1]福建师范大学数学与计算机科学学院,福州350007 [2]福建师范大学网络安全与密码技术福建省高校重点实验室,福州350007

出　　处：《南京大学学报（自然科学版）》2011年第5期493-503,共11页Journal of Nanjing University（Natural Science）

基　　金：国家自然科学基金(61070062);福建省自然科学基金2008J04004);福建省高校服务海西建设重点项目(2008HX200941-4-5);福建省高等学校新世纪优秀人才支持计划(JA11038)

摘　　要：本文基于帧间内容连续性,提出一种通过灰度值来刻画视频帧内容,利用帧间内容相关性连续度来刻画连续性与否,自适应设定阈值找出篡改点的视频篡改检测方法.首先将视频帧内容转化为一系列连续的图像帧,通过图像帧的灰度值计算帧间相关性,并计算相关性变化度,二次利用切比雪夫不等式自适应设定阈值,判断出离群点.实验表明,提出的方法检测运动背景下不小于10帧的帧删除,帧插入及帧替换篡改操作能够取得理想的效果.Recently with the development of video editing techniques it becomes increasingly easy to modify the digital video.There is an urgent need for reliable digital forensics techniques.This paper,based on interframe content continuity,proposes a digital video tampering method which adaptively sets thresholds to detect outliers.The method uses gray values to depict the video content,and uses the correlation continuity of the interframe content to describe whether video content is continuous or not.Video is an orderly time-series images in one-dimensional time domain.Therefore,for the video tamper detection,we could transfer video into series of image for testing.Grey values of frames can be a good representative of the video content,which reflect the distribution and characteristics of the image color and brightness level.Time interval between adjacent frames is small and the relevance of the content is high,but the content correlation between the far apart frames is relatively low.We calculate the correlation between the pixels in two frames to describe the continuity between frames.In order to more accurately measure the continuity of the contents of frames,in this paper,we see the absolute difference of the correlation between the frames as the basis of whether or not the video was tampered.Firstly,the video frames are translated into a series of consecutive image frames,and according to grey values of image frames,the interframe correlation is calculated.At the same time,the changing degree of correlation is computed.With the purpose of improving the efficiency,the Chebyshev inequality is used twice to set threshold adaptively and then detect outliers.The identification scheme that is proposed in this paper performs well under any widely used video formats.Experimental results show that this method can effectively detect the tempered video with moving background including frame deletion,frame insertion and frame substitution on the understanding that the tempered frames are no less than 10.

关 键 词：视频篡改 内容连续性 切比雪夫不等式 离群点检测 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]