基于奇异值优化的图像复制粘贴篡改检测算法

Detection of copy-move tampering in digital images based on improved singular value

数字图像篡改检测是数字取证领域的一个迅速发展的研究方向。复制-粘贴篡改作为一种最常见的图像篡改方式,它是通过复制一幅图像的部分内容粘贴至同幅图像另一处来隐藏或克隆对象的目的。针对已有检测算法效率不高的问题,本文提出了一种基于改进的奇异值分解的复制-粘贴篡改检测算法。该算法降低了奇异值特征向量的维度,并增加了向量相似度判别的步骤以提高准确性。实验结果表明,该算法能有效的检测复制-粘贴篡改区域,而且计算量小、检测效率高,并具有良好的检测精度和鲁棒性。

Digital image tampering detection is a burgeoning research field in digital forensics. As a most common way of image tampering, copy-move tampering is used to copy one part of the picture to another place of the same picture to conceal objects or clone regions. We reviewed several methods proposed to achieve this goal, but we found those methods are not good in efficiency. So we present an improved singular value decomposition based detection algorithm for copy-move tampering, which make the dimensional of SVD eigenvectors lower and add the step of vector similarity discrimination in order to improve the accuracy. Experimental results show this algorithm can detect the copy-move tampering regions effectively with the small amount of computation and high detection efficiency. What＇s more, it has good detection accuracy and strong robustness.