Computer Engineering and Applications ›› 2011, Vol. 47 ›› Issue (2): 185-189.DOI: 10.3778/j.issn.1002-8331.2011.02.056

• 图形、图像、模式识别 • Previous Articles     Next Articles

Image watermarking algorithm based on integer wavelet transform to resist geometric attacks

CHEN Gang1,CHEN Ning2,HU Anfeng2   

  1. 1.College of Maths and Computer Science,Jianghan University,Wuhan 430056,China
    2.School of Information Science and Engineering,Central South University,Changsha 410083,China

  • Received:2009-06-08 Revised:2009-08-05 Online:2011-01-11 Published:2011-01-11
  • Contact: CHEN Gang


陈 刚1,陈 宁2,胡安峰2   

  1. 1.江汉大学 数学与计算机学院,武汉 430056
    2.中南大学 信息科学与工程学院,长沙 410083
  • 通讯作者: 陈 刚

Abstract: Geometric attacks are difficult to resist for watermarking algorithms.Although most existing geometric-resisted algorithms can resist some kinds of geometric attacks,however,their robustness to common image processing is week.A watermarking algorithm which is both strongly robust to geometric attacks and common image processing is proposed in this paper.Watermarking messages combined a Gaussian-noise template are embedded to vertical high frequency coefficients which are produced by one-level integer wavelet transform of the host image.Then the same messages are embedded to each sub-image by quantization repeatedly.Two watermarking images are extracted by correlation calculation and majority rule.Finally,the ultimate watermarking is decided by a judge rule.Experiment results show the algorithm is both robust to geometric attacks and common image processing,meanwhile,the imperceptibility of the algorithm can meet the request.

Key words: geometric attacks, common image processing, robustness, integer wavelet transform, quantization

摘要: 几何攻击是图像水印算法较难抵御的攻击。而现有的抵抗几何攻击的算法虽然能抵抗一定程度的几何攻击,但大部分对常规图像处理攻击的鲁棒性不强。提出了一种对几何攻击和常规图像处理攻击的鲁棒性都很强的水印算法。算法在图像分块整数小波变换的垂直高频分量中结合一个高斯噪声模板嵌入水印信息,然后在空域中利用量化的方法分块重复嵌入相同的水印信息。提取时分别利用相关值计算和多数原则提取出两个水印,通过一个判决器输出最后的结果。实验结果表明,算法对几何攻击和常规图像处理攻击的鲁棒性都很强,并且不可感知性也满足要求。

关键词: 几何攻击, 常规图像处理攻击, 鲁棒性, 整数小波变换, 量化

CLC Number: