基于SIFT和非采样Contourlet变换的鲁棒水印算法

摘要

针对常见抗几何攻击水印算法存在的不足,设计一种基于SIFT和非采样Contourlet (NSCT)变换的鲁棒水印算法.算法首先利用尺度不变特征变换在载体图像的尺度空间上提取特征点并形成矩形特征区域,然后水印信息自适应量化嵌入特征区域的NSCT低频变换系数中,最后保存若干个稳定特征点用于几何校正.实验结果表明,该算法不仅具有很好的透明性,同时对常见图像处理攻击和几何攻击均具有很强的鲁棒性.%In order to solve the problem of present watermarking algorithms in resisting to geometric attacks,a robust watermarking algorithm based on scale invariant feature transform (SIFF) and nonsubsampled Contourlet transform (NSCT) is proposed in this paper.Firstly,the SIFT is adopted to extract feature points in scale space of the original image and to construct rectangular feature regions;secondly,the watermark information is embedded in the way of adaptive quantisation into the low-frequency coefficients of NSCT; finally,some of the stable feature points are reserved for geometric correction.Experimental results show that the proposed algorithm has excellent transparency and is very robust to common image processing attacks and geometric attacks.