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.%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。
展开▼
