Abstract: The existing digital watermarking is robust to add noise，filter and JPEG compression attacks，but resisting to RST geometric attacks is still a difficult challenge.It is a better method to embed the watermark into invariant regions of image.However，the embedding watermark and watermark detection are not synchronized because the image size is changed so it makes the watermark detection fail.In order to solve the problem that the size of image is changed between the embedding watermark and watermark detection，this paper proposes a novel method to resist RST transforms，which is based on model background image of zero pixels value and invariant moments.The experiment demonstrates the proposed method is better robust to add noise，filter and JPEG compression attacks，also it is better robust to RST transforms，so the proposed method is a simple，useful and credible algorithm.

Key words: robust, Rotation, Scaling, Translation（RST） transforms, invariant moments, digital watermark

摘要： 目前的数字图像水印对于加噪、滤波、JPEG压缩攻击具有很好的鲁棒性但抗几何类攻击仍是对数字图像水印的一个巨大挑战。将水印嵌入到图像中的不变域中，这是一种比较好的方法。然而，由于嵌入水印的原始图像和受到几何攻击后（RST变换）的图像尺度大小发生了改变，这就使得水印的嵌入和检测之间失去了同步，从而导致水印检测的失败。为了解决图像尺度大小在水印嵌入和检测时发生改变的问题，提出了一种使用零像素值模板背景图像和不变矩抗旋转、缩放、平移鲁棒性数字水印。通过零像素值模板背景图像代替图像尺度大小归一化，同时使用几何不变矩来设计水印和检测水印。经过仿真实验证明，该方法对于普通的加噪、滤波、JPEG压缩攻击具有很好鲁棒性的同时，对于旋转、缩放、平移等攻击也具有很好的鲁棒性，是一种简单、实用、可靠的数字水印方法（使用的仿真软件：Matlab 7.1）。

关键词: 鲁棒性, 旋转、缩放、平移（RST）变换, 不变矩, 数字水印