目的 为了有效抵抗几何攻击，实现水印图像的嵌入与检测的同步。方法 将水印嵌入到图像小波分解后的奇异值中，然后利用SIFT（scale invariant feature transform）特征点所具有的旋转、缩放和平移不变性进行宿主图像和受攻击图像的匹配，并估计受攻击图像的几何攻击参数，对可能失真的图像进行几何校正。结果 图像经过几何失真、常规图像处理攻击或JPEG压缩后，嵌入的水印依然能被可靠地检测和提取。结论 理论分析和大量实验结果表明，该算法校正精度高，具有良好的不可见性和鲁棒性。

The work aims to effectively resist geometric attack, so as to achieve the synchronization between embedding and detection of watermark image. The watermark was embedded into the singular values after wavelet decomposition. Then, SIFT (scale invariant feature transform) feature points characterized by invariance of rotation, scaling and translation were used to match host image and attacked image. The geometric distortion parameters of the attacked image were estimated and the images likely to be distorted were subject to geometric correction. The embedded watermark of the images subjected to geometric distortion, general image processing attacks or JPEG compression could still be reliably detected and extracted. Theoretical analysis and experimental results show that the algorithm has high precision and good invisibility and robustness.