提出了一种新颖的基于图像角点特征的第2代水印算法。该方法首先将水印图像进行置乱预处理，其次对载体图像进行小波3级提升，然后利用第2步的结果和多尺度曲率积角点检测（MSP）方法，提取出图像最低频成分中的角点，最后在角点位置对应的所有高频系数中嵌入水印。检测时，在这些系数上做水印嵌入的逆运算，并用事先得到的密钥反变换出水印。实验结果表明，所提出的水印算法较第1代数字水印算法不仅在运行速度上大大提高，而且在抵抗旋转、放缩、平移、裁剪等几何攻击方面的鲁棒性也有显著提高。

In this paper,a noval second generation watermarking algorithm based on image corner feature is proposed.Firstly, the watermark image is scrambled by using Arnold transformation. Secondly, wavelet transformation based three-level lifting for the carrier image is conducted. Thirdly, utilized the outcome of second step and MSP method, corners are detected from the low frequency components. Finally, by taking corners as root nodes, all the corresponding high frequency coefficients are extracted. Inverse operation of embedding watermark is performed with those coefficients. And using the given key, the watermark can be obtained. Sufficient experiments demonstrate the robustness of this approach, especially in resisting geometry attack such as RST (Rotation, Scaling and translation)and cut attack.