一种鲁棒和自适应的数字图像水印算法

为提高算法的鲁棒性和自适应能力，将原始真彩色图像的Y分量进行离散小波变换，提取出HL、LH子带的鲁棒系数嵌入水印。考虑到人眼的HVS特性，计算出HL、LH子带系数的可见误差阀值确定鲁棒系数。并根据小波分解后，同一分解阶小波系数之间的相关性，通过计算LL子带邻域均值来实现对HL、LH子带进行估计和量化，使算法的嵌入容量和鲁棒性得到增强。算法通过散列加密确定水印嵌入的位置，使攻击者即使知道算法也无法得到水印信息。实验表明算法具有在不同图像处理和攻击下，水印的鲁棒性好、检测时不需要原始图像、安全性好等优点。

A Robust and Adaptive Algorithm of Digital Water-marking

In order to improve the algorithm's robustness and adaptability, the Y component of the original real-color image is selected to carry out discrete wavelet transformation. The robust coefficients of HL and LH sub-bands are selected to embed watermarking. Considering the HVS characteristics of human eyes, the visible error thresholds are calculated for all coefficients of HL and LH sub-bands to determine whether a coefficient is robust or not. After wavelet decomposition, the wavelet coefficients of the same transformation stage are correlated. By computing the neighborhood average of the LL sub-band, the HL and LH sub-bands are estimated and quantified. Consequently, the algorithm's embedding capacity and robustness are enhanced. Since the algorithm uses hash encryption to determine the embedding position, the attackers cannot obtain the watermarking information even if the algorithm is already known. Numerical experiments prove that the proposed algorithm has advantages such as good robustness against various image processing and attacks, not requiring the original image for watermarking detection, and good security.