基于Contourlet变换与CS的彩色图像双重水印算法  

The algorithm of dual watermark in color image based on contourlet transformation and compressive sensing

作　　者：吴军[1] 孙像君 李慧慧[2] 

机构地区：[1]江西理工大学信息工程学院,江西赣州341000 [2]中国科学院深圳先进技术研究院,广东深圳518055

出　　处：《江西理工大学学报》2018年第1期71-79,共9页Journal of Jiangxi University of Science and Technology

基　　金：国家自然科学基金资助项目(61401454;71532014;U1505251);下一代通信技术国家科技重大项目(2013ZX03005013);中国科学院主要研究项目(KFZO-SW-202)

摘　　要：针对传统彩色图像水印算法抗几何攻击鲁棒性不强的问题,提出了一种基于Contourlet变换与压缩感知的彩色图像双重水印算法.该算法通过Contourlet变换与奇异值分解相结合将彩色版权水印嵌入到彩色载体图像中,然后将分块压缩感知方法应用到嵌入版权水印的彩色图像中,生成第二重水印并作为零水印进行保存.进行版权鉴定的时候,首先利用改进的平滑投影Landweber算法对零水印重构,接着使用SIFT变换原理对待检测图像进行几何校正,最后提取出版权水印验证版权.实验结果表明该算法版权水印具有较好的不可见性,除对常规攻击拥有较好的鲁棒性外,抗几何攻击能力得到了提升.In view of the weak capability of anti-geometric-attacking of the traditional watermarking algorithm,the algorithm of dual watermark in color image based on Contourlet transformation and compressive sensing（CS） is proposed in this paper. The color watermark was embedded into the color carrier image based on Contourlet transformation（CT） and Singular Value Decomposition（SVD） decomposition, and then the second watermark was generated by Block-based Compressive Sensing（BCS） from the color image embedded the copyright watermark, and was saved as zero watermark. When the copyright identification was needed, the zero watermark was reconstructed by the reconstruction algorithm of Contourlet and Smoothed Projected Landweber（Contourlet-SPL）, then Scale Invariant Features Transform（SIFT） principle was applied to correct the detection image subjected to a geometric attack, and finally, the copyright watermarking was extracted to verify the copyright. The experimental results show that the algorithm proposed hass good invisibility. In addition to the better robustness to conventional attacks, the ability to resist geometric attacks is improved.

关 键 词：CONTOURLET变换 压缩感知 奇异值分解 平滑投影Landweber算法 SIFT变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]