基于小波变换的鲁棒性与脆弱性数字水印算法的研究与实现

【作者】 郭越；

【导师】 程景云；

【作者基本信息】 上海海事大学 ， 计算机软件与理论， 2004， 硕士

【摘要】 随着计算机网络、多媒体及信息科学的不断发展，数字媒体的产权保护已成为一个迫切需要解决的问题。但是传统的信息安全技术只能控制信息传播过程，对信息失真原因判断的模糊性，对假冒欺诈的局限性，安全措施易于被觉察等弱点已经不能很好地胜任新形势的要求。作为可以解决上述问题的一种有效途径，数字水印技术开始引起人们的普遍关注。本文建立了数字水印技术中的两种典型方案，即鲁棒水印、脆弱水印，并且分别从视觉掩盖特征，水印能量，编码，密码学等方面，系统地研究了基于离散小波变换的数字水印技术，以满足各种应用的要求。 本文在上海海运学院数字图书馆工程建设的背景下，对已有水印算法进行全面的研究，提出了两种有意义的水印算法。其一，开发了一种鲁棒性水印算法“基于水印能量与HVS的自适应公开水印算法”。其二，开发了一种脆弱水印系统“具有盲重建功能的半脆弱认证水印系统”。 对于鲁棒性水印算法，提出了一种综合HVS、水印能量以及逐次逼近量化的数字水印算法。本算法没有使用加性水印的模型而是使用了Deepa Kundur的量化分段嵌入思想并对其进行了改进。通过视觉掩盖特征值来对重要系数进行量化，这样所嵌入的水印不但不会影响原图像的质量，而且还在一定程度上提高了在同样压缩率的情况下的图像质量。算法首先考虑到各种视觉掩盖效应，可以在主观感觉图像失真相同的条件下允许更大的实际失真。在分析各高频以及中高频子带的视觉掩盖特征值的基础上，有选择地嵌入水印。其次，考虑水印容量以及嵌入强度的最高阈值，结合用户所选择的PNSR值和视觉掩盖特征值来确定最佳嵌入强度。然后通过选取二值混沌序列来加密水印图像。最后采用逐次逼近量化的编码思想，这样更有利于抵抗图像的有损压缩以及计算效率。并使得在无须原始图像的情况下进行水印的提取，从而更具实用性。 对于脆弱水印系统，结合了水印技术与安全认证技术。水印标志是从原始图像小波域的稳定特征提取，通过密码学处理将水印嵌入到小波域中。在嵌入水印前对原始图像进行噪声分离以及形态学过滤处理，这样可以保证提取到的小波域特征系数对一些偶然的图像失真具有鲁棒性。本文还利用了纠错码对提取出来的特征系数进行编码，将编码得到的监督位作为水印嵌入到图像中。对图像进行认证时，可以定位被篡改位置并且恢复原图像，而真正的图像特征信息作为Hash函数的变量，由Hash函数的值产生一个全局数字签名，并应用PKI组成一个完整的水印认证框架。 本文所提到的水印算法及其试验都是在Matlab下进行。所有的试验结果表明，本文所提出的基于小波域的静态图像数字水印技术具有较好的实用性。更多还原

【Abstract】 Along with the development of computer network, multimedia and information science, problems brought by copyright protection of digital media have become more and more serious. Traditional security system of information can only safeguard information transmitting process, so they can’t be fit for requirements of new condition. As a key to solve copyright problems, watermarking technologies attract more and more attention.First, this paper introduces background of research in digital watermarking technologies; include types of digital watermarking and their principle, as well as application fields.This paper approves two kinds of wavelet-based schemes of digital watermarking. They belong to fragile watermarking and robust watermarking. The paper research these on system based on DWT in the fields of HVS, energy of image watermarking, coding and cryptography etc, which can aplly in practice.By the background of ShangHai Maritime University Digital Library Project, 1 completely study existed watermark arithmetic, approve two significative arithmetic. First, I develop a robust watermark arithmetic "Arithmetic of Adaptive Public Watermarking Based on Watermark-energy and HVS". Second, develop a semi-fragile watermark "Content Authenticate and Blind Reconstruct Based Semi-Fragile Watermark System".With regards to robust arithmetic, an embedded-coding watermark algorism based on HVS and watermarking energy is presented. This algorism embeds watermark in key coefficients of wavelet domain of images, and it provides the mechanism of embedding the optimal intensity of watermark for users according to the PNSR and visual-masking eigenvalue given by users. The paper also employs and improves the idea of quantitive subsection embedding in the implementation of embedding watermark. Experiments results show that the method offers fine invisibility and robustness in image processing.Regarding to semi-fragile arithmetic. We introduces the technique of noise separation to deal with the wavelet coefficients, and uses the morphological theory to screen and select the characteristic coefficients, which ensure the robustness of the above mentioned coefficients. On the base of the method, we encode the selected characteristic coefficients with the error correction code, in the meantime embeds into images the error correction bits that are subject to the processing of dispersal. Authentication of image content using PKI can locate actual places of the tampering and rehab the image, but also fortify the security and integrity of the system.This paper’s watermarking arithmetics are all run in Matlab environment. All the experimentation result show that this paper’s static image watermark based on wavelet potential application promises well.更多还原