Abstract: The image watermark scheme based on Non-Subsampled Contourlet Transform(NSCT) is proposed for resisting Rotation Scale Translation(RST) attack. Feature points are extracted by utilizing the characteristic of image from the greatly directional band pass coefficients of the NSCT processed by binary. When watermark is detected, the feature points matching measure for the geometric manipulation is computed. Compared with other algorithm, the method can transform the image back to its original size/orientation, i.e, recover the lost synchronism. Experimental results show that it is robust to common geometric attacks.

Key words: Non-Subsampled Contourlet Transform(NSCT), contour directional coefficients, digital watermark, binary processing

摘要： 提出一种利用非采样Contourlet边缘方向信息实现对RST攻击校正的水印方案。该方案利用非采样Contourlet变换提取载体图像多尺度多方向上的变换系数，对显著方向子带系数进行二值并细化，获得载体图像轮廓方向信息。水印检测时，利用轮廓方向的特征点集估计测试图像受到RST攻击的参数，从而使测试图像恢复到原来的大小和位置，即恢复已丢失的同步信息。实验结果表明，该方法对RST攻击具有较好的鲁棒性。

关键词: 非采样Contourlet变换, 轮廓方向系数, 数字水印, 二值化

CLC Number: 

• TP309