抗JPEG攻击的图像水印算法研究

【作者】 罗大光；

【导师】 郝玉洁；

【作者基本信息】 电子科技大学 ， 计算机应用， 2005， 硕士

【摘要】 随着数字网络通讯的飞速发展以及数字多媒体的广泛应用,对数字产品版权保护的需求也日益迫切。数字水印技术已经成为用来解决数字多媒体中版权问题和内容认证的主要工具。本文首先介绍了数字水印的研究背景及国内外研究动态,描述了数字水印的基本特征、分类及其应用。接着系统地论述了图像水印算法的通用模型,并对其中的各个主要组成部分进行了详细的阐述。本文主要对基于图像的数字水印算法进行了研究,并基于空间域和变换域分别提出了两种新的图像水印算法,主要内容如下: 一种是基于图像最高有效位(MSB)的数字水印算法:数字水印技术的研究是从空间域开始的,最低有效位(LSB)算法是空间域算法中最有代表性的,它的缺点主要是嵌入位置不可靠,鲁棒性非常差,特别是对JPEG 压缩攻击毫无抵抗之力。另一方面,针对图像最高有效位(MSB)的研究文献却比较少见,这是因为MSB 包含了颜色的最主要信息,对其进行修改容易引起图像质量明显的下降,违背了水印不可见性的条件。我们提出了在MSB 中嵌入水印的算法,通过选择使用一些特定的颜色分量值,实现了对图像MSB 的利用。该算法是一个非盲检测算法,在检测过程中需要原始图像的参与。实验结果表明,该算法对JPEG 压缩攻击具有一定的鲁棒性。一种是基于层次DCT 的盲水印算法:与国际压缩标准的兼容性使得DCT 域的水印算法可以很好地利用压缩域的特点进行水印的嵌入,水印的鲁棒性也较强,并且计算简单,因而出现了大量基于DCT 域的水印算法的研究。研究发现,DCT 能产生类似于小波的多分辨率图像分解特性,而这种特性有利于人们将人眼视觉特性与图像的统计特性结合起来。因此,我们提出了一种基于层次DCT 的图像水印算法,目标就是要在DCT 域中利用这些特性,进一步改善图像水印算法的性能。该算法实现的是盲检测,在检测过程中不需要原始图像的参与,并且提供了主、客观两种不同检测方式。实验结果表明,该算法对JPEG 压缩攻击、剪切攻击有良好的鲁棒性。更多还原

【Abstract】 With the rapid growth of Internet technologies and extensive use of multimedia data, the enforcement of multimedia copyright protection becomes an important issue. Digital watermarking is viewed as an effective way for intellectual property right protection of multimedia data. Firstly, This paper introduces the development of digital watermarking technologies, describes the basic characteristic, classification and application fields of digital watermarking. It then formulates the generic model of digital watermarking algorithms based on images and expatiates the main parts of the model. This paper is focused on images watermarking algorithms. Two new algorithms based on spatial-domain and transform-domain respectively are presented as follow: A digital watermarking algorithm based on Most Significance Bits (MSB) of images is proposed. As few digital watermarking algorithms based on spatial-domain can survive after familiar attack, researchers are no more interested in them as ever. One of the algorithms is based on Least Significance Bits (LSB), which is not good at embedding space and robustness, especially unable to resist JPEG compressing. Considering the change of MSB often greatly affects the quality of images, few researchers turn their attention to MSB. That’s why few papers about MSB can be found now. Here a watermarking algorithm based on MSB is proposed. It works by using some special color partial-value. In addition, the original image is needed in detecting phase. Pilot study shows that the proposed algorithm has some robustness to JPEG compression. A blind digital watermarking algorithm based on Hierarchical Discrete Cosine Transform (HDCT) is proposed. Digital watermarking algorithm based on Discrete Cosine Transform (DCT) transform-domain has been widely studied for DCT is used in international compressing standard and is helpful to create robust watermarking algorithms. Research shows that the reorganization of DCT can be taken as multiresolution wavelet decomposition, which can combine the characteristics of human vision and更多还原