基于Watson模型的改进QIM盲水印算法  

Blind QIM watermarking algorithm based on the Watson model

作　　者：胡自权[1] 佘堃[1] 刘金华[1] 

机构地区：[1]电子科技大学计算机科学与工程学院,成都611731

出　　处：《电子测量与仪器学报》2010年第9期824-830,共7页Journal of Electronic Measurement and Instrumentation

摘　　要：2007年,LiQ.和CoxIJ.提出了一种基于离散余弦变换的自适应QIM算法。该算法为保存水印图像需在嵌入水印后对值进行四舍五入取整,造成误差最大达0.5。针对这一问题,利用小波变换的特点,结合Watson感知模型,提出了一种改进的QIM盲水印算法。该算法使用小波分解原始图像,利用Watson感知模型计算并自适应选取量化步长,生成重构图像,微调重构图像,产生水印图像,在小波域中值的误差最大0.25。该算法提取水印时,只依赖Watson模型重新计算的量化步长,不依赖原始图像或原始水印信息,该算法是盲检测算法。实验仿真结果表明:改进算法在抗加性高斯白噪声攻击、JPEG压缩攻击、中心裁剪攻击、中值滤波和伸缩比因子变化等方面都比原算法有更好的性能。In 2007,an adaptive QIM method based on Discrete Cosine Transform was proposed by Li Q and CoxI J.The algorithm needs to round the value for saving the watermark image,so the result of the algorithm may lead tothe error that is 0.5.Applying the wavelet transform,the blind QIM watermarking algorithm based on Watson model isproposed in this paper.In the new algorithm,after decomposing the indexed image of the original image,the step sizeof the quantizer via Watson model is computed and the image is reconstructed.Through minute adjustment of the reconstructed image value,the algorithm obtains the watermarked image,and its error is 0.25 in the wavelet domain.Because the watermark message is extracted not by the original image but by computing the step size of the watermarkedimage,the algorithm is the blind watermarking method.The experimental results with comparison the method show thatthe improved algorithm has superior performance to the original one in the aspect of resistance to additive white Gaussian noise,JPEG compression and so on.

关 键 词：数字水印 Watson感知模型 小波变换 鲁棒性 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]