摘要
提出一种新的基于核判别分析的图像水印方法.载体图像被划分为不重叠的图像块,并且每个图像块被进一步分为两个子图像块.所有这些子图像块分别进行小波分解.利用图像块的两个子图像块局部能量之间的某种关系来隐藏水印信息.为了提取水印,核判别分析方法被用作分类器来学习这种关系.通过利用核判别分析良好的学习能力,水印能在几种不同攻击下被正确地提取,并且不要求原始图像参与.实验结果表明本文方法在对抗诸如JPEG压缩、加噪、低通滤波、锐化等攻击有比其它几种方法更好的性能和鲁棒性.
A new image watermarking method based on kernel discriminant analysis is proposed in this paper.The host image is partitioned into nonoverlapping image blocks,and then every image block is divided into two sub-blocks.All sub-blocks are decomposed with wavelets transform respectively.The watermark information is embedded into the low-frequency sub-bands by applying some relationship of local energy between two sub-blocks of image block.To extract the watermark,the kernel discriminant analysis is used as the classifier to learn the relationship.By applying the good learning ability of kernel discriminant analysis,the watermark can be correctly extracted under several different attacks,and the original image not is required.Experimental results show that the proposed method has good performance and robustness to resist common attacks such as JPEG,addition noise,low pass filtering,blurring,etc,which is superior to other several methods.
出处
《小型微型计算机系统》
CSCD
北大核心
2010年第12期2491-2496,共6页
Journal of Chinese Computer Systems
基金
四川省教育厅重点项目(07ZA112)资助
四川省学术和技术带头人培养资金项目(08209057)资助
西华大学重点项目(ZG0722603)资助
西华大学重点实验室基金项目(XDZ0818-09)资助
关键词
数字水印
图像水印
核判别分析
digital watermarking
image watermarking
kernel discriminant analysis