基于重采样痕迹的图像伪造检测

摘要

检测重采样痕迹是数字取证中判断图像是否被篡改的有效途径之一。针对现有重采样检测方法大多只考虑单次重采样情况，对再次经历重采样的伪造图像不能有效区分定位篡改区域这一问题，提出一种基于重采样痕迹的图像伪造检测算法。首先定义出能够描述并区分不同重采样痕迹的两个特征量，将待测图像重叠分块，计算每块的特征量，然后利用特征量的不一致性检测定位篡改区域。实验结果表明，该方法能够区分旋转与缩放的操作历史痕迹，进行篡改伪造图像的自动判断与篡改区域定位；并且当伪造图像再次经历重采样操作后，仍能区分出图像中的不同插值区域，即对再次重采样操作具有一定的鲁棒性。%To detect resampling traces is one of the effective ways in digital forensics to determine whether or not the image has been tampered.Existing resampling detection methods mostly consider the single resampling cases only but cannot effectively differentiate and locate the tampered region on the forged image with secondary resampling.Aiming at this problem,we propose in this paper a resampling trace-based algorithm for image forgery detection.We first give the definition on two feature values capable of describing and differentiating different resampling traces,and then overlap and divide into blocks the measuring images to calculate the feature value of each block,and use the inconsistency of feature values to locate the tampered regions.Experimental results show that the algorithm is able to distinguish historical traces of rotation and scaling operations,to make automatic determination on tampered and forged images and localisation of tampering region. Moreover,when the forged image comes through resampling operation once again,it can still differentiate different interpolation regions in the image,i.e.it has certain robustness on secondary resampling operation.