Many block-based detection methods for image copy-move forgery have been reported. However, their performance degrades significantly under different geometric attacks such as rotation and scaling. In this paper, we propose a novel robust and accurate detection scheme for image copy-move forgery. It mainly consists of three steps: firstly, a suspicious image is divided into overlapping circular blocks, and polar complex exponential transform (PCET) is employed to extract geometric invariant feature of each block. Next, singular value decomposition (SVD) is applied to the coefficient matrix composed of extracted geometric invariant moments for dimension reduction. Meanwhile, the histogram of block similarity measures is adopted to estimate the optimal similarity threshold. Finally, the calculated similarity threshold is used for block matching process and consequently more accurate tampered areas are obtained. Experimental results on various datasets show that the proposed image copy-move detection approach outperforms other existing methods in the aspect of resisting geometric rotation and scaling attacks, with the adaptability of similarity threshold and low computational complexity.

已经报道了许多基于块的图像复制移动伪造检测方法。但是，它们的性能在不同的几何攻击（例如旋转和缩放）下会大大降低。在本文中，我们提出了一种新颖的鲁棒且准确的图像复制移动伪造检测方案。它主要包括三个步骤：首先，将可疑图像划分为重叠的圆形块，然后采用极复指数变换（PCET）提取每个块的几何不变性特征。接下来，将奇异值分解（SVD）应用于由提取的几何不变矩组成的系数矩阵，以进行降维。同时，采用块相似性度量的直方图来估计最佳相似性阈值。最后，计算出的相似度阈值用于块匹配过程，从而获得更准确的篡改区域。在各种数据集上的实验结果表明，所提出的图像复制移动检测方法在抵抗几何旋转和缩放攻击方面优于其他现有方法，具有相似阈值的适应性和低计算复杂度。