【作者】 侯晓芳；

【导师】 李东生；

【作者基本信息】 太原理工大学 ， 计算机应用技术， 2007， 硕士

【摘要】 随着数字通信网络技术与多媒体技术的迅速发展，越来越多的多媒体作品通过网络的形式发布，使信息的发布和传输变得方便快捷。但数字作品具有极易被非法复制和篡改的特性，任何人都可以免费并且没有任何质量损失地批量复制和发行受知识产权保护的内容，这对数字产品的版权保护造成了严重的威胁。近年来，数字产品的版权纠纷案件越来越多，侵权盗版活动也日益猖獗，大范围的侵权拷贝行为受到了音像、影视、出版等行业的高度重视。于是，数字水印技术孕育而生，它作为传统密码学的有利补充，为实现有效的信息版权保护手段提供了一条崭新的思路，成为多媒体信息安全研究领域的一个热点，在信息安全问题、盗版和版权纷争问题、多媒体内容的版权保护与内容鉴别等方面发挥着它的作用。本文主要研究了变换域中基于DCT和DWT的两种图像数字水印方法。在水印算法的设计中综合了影响水印性能的各种因素，并针对不同的变换域，重点分析了水印嵌入策略中水印嵌入位置对水印性能的影响。文中实现了一种有意义的基于DCT的公开水印算法，并引入随机序列控制嵌入位置间隔的思想对算法进行了改进。算法采用混沌序列作为随机序列的发生器，参照压缩亮度量化值选择系数，并选取控制量进行系数调制，使水印系统具有较好的性能。改进后的算法与改进前相比较，水印系统的安全性得到了提高，且对抗剪切、JPEG压缩等攻击的能力明显增强。另外，水印的嵌入不使用传统的水印嵌入模式，避免了传统水印嵌入模式下的模型攻击问题。水印信号的提取过程不需要借助原始图像，满足水印盲提取的要求。论文结合新的图像压缩标准JPEG2000和HVS理论，实现了一种基于DWT的图像数字水印方法。该算法依据小波变换的多分辨率特性，选取高尺度分解下的高频系数来嵌入水印，这样更有利于满足水印系统对鲁棒性和不可见性的要求，体现出小波域数字水印的优势。算法通过对DWT系数进行分析和大量的实验验证，设定了较为合适的两个阈值来控制系数的选取。该算法水印的嵌入与图像的统计特性相关，水印系统能够自适应于图像。更多还原

【Abstract】 With the rapid development of digital communication network and multimedia technology, more and more multimedia products have been issued through network form, which also have made the issuance and transfer of information become not only convenient but also quick. But there are properties such as being copied and interpolated illegally in digital products. Anybody can copy and issue what is protected by the intellectual property rights in batches for free and without any quality loss, which have critically threatened the copyright protection of digital products. In recent years, there are more and more copyright dispute cases of digital products and the act of tort become rampant day by day. Right infringements of copy behaviors have been laid store by the industry of video, film and publication. So the digital watermark technology which is regarded as the powerful complementarily of the traditional cryptology was born. The digital watermark technology provides a brand-new idea for realizing effective information copyright protection method and it also becomes a hot topic in the research area of multimedia information security. It is playing a good role in questions of information security, pirate and copyright dispute, copyright protection and content distinction of multimedia, and so on.Two kinds of digital watermark methods about image based on DCT and DWT in frequency domain were mainly researched in this article. In the design of the watermark algorithms, all kinds of factors affecting watermark performance were synthesized and the influence of position watermark embedded to watermark performance at different frequency domain was analyzed with emphasis.In this article a meaningful public watermark algorithm based on DCT was realized. An idea of controlling the position embedded by the random series was inducted which improved the algorithm. Algorithm uses Logistic chaotic series as the random series generator, selects coefficients according to the luminance quantization value of compression and chooses controlled variable to do the coefficients modulation, which will make the watermark system be of good performance. Compared to the original algorithm, the security and the ability of resisting some attacks such as cutting and JPEG compression were obviously enhanced in the watermark system. Otherwise, the embedded mode of traditional watermark doesn’t been adopted, which avoids the model attack problems under the traditional watermark. The extract process of watermark signal doesn’t have to resort to the original image to satisfy with the requirement of blind extract. A kind of digital watermark method about image based on DWT was realized in this paper, combining the JPEG2000 which a new image compression standard and HVS. The high frequency coefficients under high scale decomposition according to multi-resolution characteristic of wavelet transfer were chosen in the algorithm, which is advantageous to satisfy the requirements of watermark system to robustness and invisible and manifests the advantages of digital watermark in wavelet domain. It sets two more appropriate threshold values to control the selection of the coefficients through analyzing DWT coefficients and being confirmed by massive experiments. The watermark embedded in this algorithm is concerned with the characteristic of image statistic and the watermark system can self-adapt to the image.更多还原