基于人类视觉系统的自适应水印嵌入方法  

作　　者：王秀丽[1] 谢卫华[2] 王翾[3] 

机构地区：[1]中国传媒大学信息工程学院,北京100024 [2]中国传媒大学计算机学院,北京100024 [3]中国传媒大学广播电视数字化教育部工程中心,北京100024

出　　处：《中国电子商情（通信市场）》2012年第3期71-78,共8页

基　　金：致谢本论文受国家863计划“面向三网融合的集成播控平台研究与示范”（2011AA01A107）项目资金资助.

摘　　要：为保证数字图像水印的鲁棒性和不可见性，本文提出了一种基于DCT域DC系数二次变换的水印嵌入算法。该算法首先将原始载体图像进行一次分块DCT变换，提取一次DCT变换后的DC系数，将其生成新矩阵后进行二次DCT变换，通过多种实验结果比较，选择将水印嵌入到DC系数二次变换后的高频系数上，并在此基础上，结合人眼视觉系统（HVS）的亮度掩蔽特性以及不同的块可嵌入的水印容量的不同，通过计算二次DC系数矩阵每个分块的方差，自适应选择可嵌入水印的DCT块。该算法不仅具有较好的不可感知性，而且对诸如JPEG压缩、旋转、几何剪切、叠加噪声等攻击均具有较好的鲁棒性，且能保证含有水印的图像的质量。Considering the robustness and imperceptibility of digital image watermarking, this paper proposes an watermark embedding algorithm, in which watermark is embedded in DC coefficients derived from a dual DCT transformations. After block based DCT transformation, DC coefficients of an image are arranged into a new matrix, on which block based DCT transformation is performed again. Through abound empirical tests, high frequency coefficients of DCT transformation are selected to be embedded with watermark. Allowing for properties of the human visual system （HVS）, variance of each block is calculated to determine whether a bit will be embedded into it. Simulation tests demonstrate that the algorithm not only has good imperceptibility, but also has good robustness for JPEG compression, rotation, geometric shear, imposed noise and other attacks, etc.

关 键 词：数字图像水印DCT变换 直流系数(DC) 二次变换 

分 类 号：TN911.73[电子电信—通信与信息系统]