设计抵抗RST(旋转、缩放、平移)攻击的数字水印算法是数字水印技术研究中的难点。针对这个难点,本文采用了广泛应用于模式识别中的几何矩技术,利用3个原始图像的几何矩和遭受攻击后图像的几何矩信息准确地估计出水印图像所经历的几何攻击,然后对其逆变换校正,确保了水印嵌入与检测的同步,极大的提高了水印算法抵抗RST攻击的能力。本文实验结果表明:该算法对抗RST攻击具有很好的鲁棒性。
Resistance to rotation, scaling, translation (RST) attacks is a difficulty of digital watermarks. A novel watermark algorithm which is resistant to geometric attacks based on geometric moment was proposed. Geometric distortions of the watermarked image were estimated accurately and its inverse transformation was corrected by geometric moments, of the original and the watermarked images, used for pattern recognition, to ensure synchronization between embedding and detecting of the watermark and improve the resistance to RST attacks. The simulated results demonstrate that the algorithm is very robust to RST attacks.