采用高斯几何矩的图像复制粘贴篡改盲检测

An algorithm for detecting image copy-move forgery regions using Gaussian geometric moments

针对图像盲检测中一种常见的简单的篡改方式——复制粘贴,提出了采用高斯几何矩的图像篡改检测方法.首先将图像进行重叠分块,其中所分块的大小相同,然后提取每个图像块的高斯几何矩作为对应的特征向量,通过字典排序之后,计算图像块间的欧式距离,最后根据图像块的相似性确定篡改区域.实验结果表明,该方法能够有效地检测与定位图像的篡改区域,与不变矩相比有较强的抗噪性,能够成功地检测出带有噪音的篡改图像,同时具有很好的鲁棒性.

Copy-move is a common digital image forgery in the research field of image blind authentication. A Gaussian geometric moment based algorithm for detecting and locating the copy-move forgery region in one image is proposed in this paper.In the algorithm,the eigenvectors of image are produced by extracting the Gaussian geometric moment features of every image block.Then the eigenvectors of image are sorted by dictionary ordering. The similarity between the image blocks is computed by Euclidean distance for detecting and locating the forged image region.The experimental results indicate that the method can effectively detect and locate the tampered im- age region.Compared with the invariant moments, the algorithm has the stronger noise resistance and can successfully detect the tampered image with noise by good robustness.