中文摘要：

几何攻击作用在隐秘图像上，这导致检测水印失败，除非在检测水印之前能使受到几何攻击的隐秘图像恢复到与原始图像具有同样大小和位置，也就是恢复已失去的同步信息．为了做到这一点，解码器需要知道隐秘图像受到了怎样的几何攻击．文章提出了一种新的估计几何变换的方法，它利用图像的形殊点来实现这个目的．分别计算出原始图像和测试图像的形殊点集，通过模拟退火算法对两个形殊点集进行匹配，从而估计出几何变换的参数．所提出的方法在检测水印时不需要原始图像做参考，这是因为原始图像的形殊点信息已包含在密钥中．实验结果表明该方法对几何攻击和其它常用图像处理攻击具有鲁棒性．

英文摘要：

Geometric attacks performed on a watermarked image, as part of the attacks the image may undergo, can very easily confuse the decoder unless it transforms the image back to its original size/orientation, i. e. , recover the lost synchronism. To be able do so, the decoder needs to know how the watermarked image has been manipulated. A novel method to estimate the geometric manipulation is reported in this paper. The feature points matching measure for the geometric manipulation is computed. The reference points of the original image are computed using shapespecific points （SSPs）. The feature point set of the tested image is obtained using SSPs, too. The feature points matching are realized by simulated annealing algorithm. The proposed scheme, when watermarking is detected, does not require the original image because reference point information of the original image has been contained in the secret key. The experimental resuits show that its robustness to geometric attacks, and other most common image processing manipulations and attacks.