基于对象的图像篡改盲取证研究  

Blind evidence tampering object-based image

作　　者：陈渝[1] 琚生根[2] 夏欣[2] 何坤[2] 

机构地区：[1]四川民族学院计算机科学系,康定626001 [2]四川大学计算机学院,成都610064

出　　处：《四川大学学报（自然科学版）》2014年第3期495-498,共4页Journal of Sichuan University(Natural Science Edition)

基　　金：四川省教育厅自然科学项目(13ZA0135);四川省科技支撑计划项目(2012GZ0091);四川大学青年基金项目(2012SCU11033)

摘　　要：根据自然界不存在完全一致的景物,提出了图像对象复制粘贴篡改的盲取证算法.分析了对象亮度和几何变化等因素对盲取证影响;运用灰度归一化和不变矩提取各个对象的盲取证特征;最后分析各个对象的相关性,若其相关性大于某一阈值则表示相同的景物,图像中存在对象复制粘贴篡改.实验结果表明,本文取证算法在一定程度上克服了对象亮度和几何变形对认证的影响.Based on the nature does not exist exactly the same scene, this paper presents an image object copy and paste tampering forensics algorithm. Firstly, the object brightness and geometry changes and other factors on forensics. Secondly, the use of gray-scale normalization and invariant moments to ex- tract features of each object forensics, Finally, analysis of the correlation of each object, if its relevance is greater than a certain threshold, said the same scene, the presence of an object image copy and paste tampering. Experimental results show that the proposed algorithm forensics overcome to some extent subject brightness and geometric deformation on certification.

关 键 词：图像对象 复制粘贴 ZERNIKE矩 篡改认证 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]