一种安全多功能图像水印系统模型与算法  

A Security Multipurpose Image Watermarking System Model and Algorithm

作　　者：朱从旭[1] 陈志刚[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《中国图象图形学报》2008年第5期894-899,共6页Journal of Image and Graphics

基　　金：国家自然科学基金项目(60573127);教育部博士点专项科研基金资助项目(20040533036)

摘　　要：为提出一种安全多功能的图像水印系统,首先利用3维统一混沌系统的混沌序列Y、Z生成可见水印的嵌入系数,同时在图像的空域嵌入可见水印,以实现版权通知功能;然后在含有可见水印的图像中加入第1种不可见水印,以实现版权保护功能;最后再在消除可见水印的图像中加入第2种不可见水印,以实现非授权拷贝跟踪。两种不可见水印的嵌入位置都由统一混沌系统的混沌序列X、Y随机选择,并基于一种新的双参数量化技术在选择的图像子块小波域低频系数中嵌入水印,并能进行盲提取。实验结果表明,可见水印能被授权用户完全消除,但很难被非授权用户消除;而不可见水印对图像的质量影响很小,并对常规图像处理操作具有良好的鲁棒性。To develop a security multipurpose image watermarking system,we use the 3D unified chaotic system sequences Y and Z to generate embedding coefficients of visible watermarks, and embed the visible watermarks in the image spatial domain to achieve the goal of copyright notification. Then we embed the first invisible watermarks in the image with embedded visible watermarks to obtain the aim of copyright protection. After removing the visible watermarks, the second invisible watermarks are embedded in the image to obtain the aim of unauthorized copy tracing. Two kinds of invisible watermarks＇ embedding positions are selected randomly by unified chaotic system sequences X and Y, based on a novel dual parameters quantization technique. Both watermarks that can be blindly extracted are embedded in the lifting wavelet transform （LWT） domain of those selected boeks. Simulation results show that the visible watermarks can be removed entirely by authorized users and can hardly be removed by unauthorized users. The impact of the invisible watermarks on the image quality is slight, and the invisible watermarks are robust to common image processing operations .

关 键 词：水印模型 提升小波变换 统一混沌系统 可见水印 不可见水印 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]