摘要
目前,大多数鲁棒图像水印所面临的最大问题就是几何攻击,而几何攻击又以仿射变换为主,攻击者只需对目标图像做轻微的几何形变就可以使嵌入其中的水印信息失效。本文在Ping Dong等人提出的基于图像归一化方法的抵抗任意仿射交换攻击的鲁棒数字水印的基础上,改进了原有的归一化方法,去掉了归一化过程中的约束条件,简化了归一化的过程。并且在此基础上提出一种新的水印嵌入方案,克服了文献[1]中水印方案水印不可见性较差的问题。实验证明,本文提出的方案对仿射变换攻击具有很好的鲁棒性,嵌入的水印容量较大且水印图像的质量较高。
Geometric attacks are the challenging problems that most robust image watermarking systems face today. Most
Geometric attacks are affine transformations, and even imperceptible geometric transformation can defeat the detection process
although the watermarking information still exists in the image. A modified image normalization method and a new multi-bits
watermarking imbedding algorithm are proposed in this paper based on the scheme resistant to affine transformation attacks
proposed by Ping Dong. Image normalization process is simplified and a new multi-bit watermarking embedding scheme is
proposed to overcome the disadvantage of low imperceptibility. Experiments indicate that the scheme proposed in this paper is
very robust to affine transformation, all has large capacity and good image quality.
出处
《信号处理》
CSCD
2004年第3期245-250,共6页
Journal of Signal Processing
基金
国家自然科学基金(No.60172062
No.60373028)
霍英东青年教师基金(No.81053)
留学归国人员科研启动基金