基于Zernike矩的图像区域旋转篡改检测

Detection by Region Rotation Duplication Based on Zernike Moments

区域旋转复制粘贴是极为常见的图像篡改方式之一,为了有效地检测这种图像篡改,该文提出一种基于Zernike矩旋转不变性的篡改检测算法。首先,将图像划分为相互重叠的小块并计算各小块的Zernike矩特征值,对其进行字典式向量排序;其次,计算字典式矩阵相邻行之间的相关系数确定旋转篡改区域;最后,估计区域的旋转角度。实验结果表明,该算法不仅能够精确定位图像中旋转篡改的区域,而且能估算出区域旋转角度。可用于面向图像内容的真实性认证。

The region rotation copy-move is a common method in digital image manipulation,based on Zernike moments, a detection algorithm of rotation invariance is proposed. Firstly, the image is processed with block-based Zernike. Then the Zernike moments vectors are lexicographically sorted and duplicated image blocks will be close in the sorted list, and therefore the correlation of neighbor rows is calculatcd during the detection steps. Experimental results demonstrate that the proposed approach can not only localize the duplicated regions accurately ,but also evaluate the rotation angle effectively. The method suits to the forensics for image contents.