中文摘要：

图像的复制-粘贴篡改是常见的图像篡改方法之一.现有基于SIFT特征的算法能够有效地检测复制-粘贴篡改,但由于SIFT特征本身不能抵抗翻转,因此,这些方法不能检测出具有翻转操作的复制-粘贴篡改.基于SIFT特征,提出了一种抗翻转的图像复制-粘贴篡改检测算法.通过在检测框架中引入图像预处理操作,不仅能够有效地检测出存在翻转的复制-粘贴篡改块,而且能够抵抗旋转、缩放等图像处理行为.同时,在SIFT关键点匹配环节提出了ng2NN匹配方法,提高了算法的检测效果.实验结果证明了所提出算法在抵抗翻转、缩放、旋转以及检测多重复制-粘贴篡改等方面的有效性.

英文摘要：

The copy-paste is one of common methods of image tampering.Although the existing SIFT-based algorithms of copy-paste detection work well facing typical copy-paste tampering, they often have limited capability to detect copy-paste tampering with flip operation because SIFT is not flip invariant. Based on SIFT, an anti-flip detection algorithm is proposed. By introducing image pre-processing into the detection framework, the proposed method can detect copy-paste not only with flip processing, but also with rotation and scaling.At the same time,the ng2NN matching method is proposed in the keypoint matching step to increase the detection effect.The experimental results show that the proposed method can he effective to resist flip operation, scaling and rotation, and to detect multiple copy-paste tampering.