摘要
为了进一步提高水印算法的鲁棒性,提出了一种在无下采样Contourlet变换(non-subsampled contourlet transform,NSCT)域中利用尺度不变特征变换(scale-invariant feature transform,SIFT)特征点进行几何攻击校正,并结合非负矩阵分解(non-negative matrix factorization,NMF)和奇异值分解(singular value decomposition,SVD)的图像抗几何攻击水印算法。该算法首先对RGB宿主图像进行两层NSCT并分别提取低频部分的红色分量和蓝色分量;然后充分利用非负矩阵的线性无关性和稀疏性以及奇异矩阵的稳定性,对蓝色低频分量进行水印的嵌入;最后利用红色低频分量的SIFT特征点信息对宿主图像进行几何攻击校正,恢复水印的同步信息后再提取水印。大量的实验结果表明,该算法在保证不可感知性的前提下,对于常规图像处理具有更好的鲁棒性能,并能有效地抵抗各类几何攻击和组合攻击。
In order to further improve the robustness of watermarking scheme,an image watermarking scheme robust to geometric attacks in non-subsampled contourlet transform( NSCT) domain is proposed,in which scale-invariant feature transform( SIFT) key-points are used to correct the geometric attacks,and non-negative matrix factorization( NMF) is combined with singular value decomposition( SVD). Firstly,two-level NSCT of RGB host image is performed. The red component and blue component of the low-frequency part are extracted,respectively.Then the linear independence and sparsity of the non-negative matrix after NMF,and the stability of the singular matrix after SVD are made full use of. The watermarking is embedded into the blue low-frequency component. Finally,the SIFT key-points of the red low-frequency component are used to correct the geometric attacks for host image,the watermarking synchronization information is recovered and the watermarking is extracted. A large number of experimental results show that,on the premise of invisibility,the proposed algorithm is more robust against the common image processing,while it can effectively resist the geometric attacks and the combination attacks.
出处
《电子测量与仪器学报》
CSCD
北大核心
2015年第7期961-969,共9页
Journal of Electronic Measurement and Instrumentation
基金
国家自然科学基金(60872065)
江苏省社会安全图像与视频理解重点实验室(南京理工大学)开放基金(JSKL201302)
江苏高校优势学科建设工程资助项目
关键词
数字水印
无下采样Contourlet变换
非负矩阵分解
奇异值分解
尺度不变特征变换
digital watermarking
non-subsampled contourlet transform
non-negative matrix factorization
singular value decomposition
scale-invariant feature transform
