【作者】 石慧；

【导师】 王相海；

【作者基本信息】 辽宁师范大学 ， 计算机应用技术， 2008， 硕士

【摘要】 近年来,数字水印技术得到了飞快的发展,目前抗几何攻击成为水印研究的热点和难点问题,据此本文以静态图像为研究对象,提出了两种抵抗几何攻击的数字水印算法。此外,本文还提出了一种半脆弱数字水印算法,半脆弱水印与抗几何攻击水印要求相反,它只要对常见的图像处理具有一定的鲁棒性即可,但半脆弱数字水印要求所嵌入的水印信息能够敏感地检测并定位对含水印图像进行的非法篡改。本文所提出的三种数字水印算法充分利用了正交变换的特性,分别利用离散余弦变换(DCT)、提升小波变换、Contourlet变换三种工具提出了以下三种水印算法:1.一种基于DCT域的半脆弱数字水印算法。该算法利用DCT的分频特性,结合HVS的掩蔽特性,将嵌入的水印能量较好的分布到图像块的各部分,保证了所嵌入的水印在不可见性的前提下具有很强的鲁棒性,并且具有一定的检测篡改的能力,可以定位到哪个块的系数被改变。2.一种基于边信息及仿射模型的抗几何攻击的数字水印算法。该算法提取小波变换近似子带中的边框系数作为边信息,利用该边信息及仿射模型估计各种几何攻击,在此基础上提出了一种抗几何攻击的小波域自适应图像水印算法,该算法根据HVS的掩蔽特性将水印信息自适应地嵌入图像小波域的低频子带中,算法不仅具有较好的不可感知性,而且对诸如平移、旋转、缩放、镜像和转置等几何攻击及其它常规的图像攻击都具有较强的鲁棒性。3.一种基于Contourlet变换的抗几何攻击数字水印算法。它首先利用Contourlet变换中不同分辨率下的父子关系,提出了空间方向重要树的概念,据此确定重要系数并嵌入水印信息。同时利用归一化方法,通过计算比较受攻击前后图像的特征量来恢复图像,提取水印信息。试验结果表明,该算法对于水平及垂直翻转攻击、旋转攻击、尺度变换攻击等具有较好的鲁棒性。更多还原

【Abstract】 Recently digital watermarking gets the rapid development, presently robust to geometric attacks become the hotspot and difficult problem in the research of watermarking.Hereby this text makes the still picture as study object, and put forwards two kinds of watermarking algorithm which robust to geometric attacks. Besides, this text also put forwards a semi-fragile watermarking.The request of semi-fragile watermarking and robust to geometric attacks’watermarking is oppositive, if only the semi-fragile watermarking has a little robustness to common image process, but it must has the capability of checking and locating the unlawful tamper in watermark image. The three kinds of watermarking algorithm make best used of orthogonal transform’s character, according to Discrete Cosine Transfor, Lifting Wavelet Transform and Contourlet Transform to put forward this three watermarking algorithm:1. A Semi-fragile Watermarking Algorithm Based on DCT. This algorithm uses DCT merit, and combine the HVS masking character, it makes the built-in watermarking enery distribute the image blocks preferably .In the promise of imperceptible, it has strong robustness.Also it has capability of checking tamper, it can orient 8*8 block in the spatial domain.2. A Robust to Geometric Attack’s Watermarking Based on Edge Information and Affine Model. It makes the frame coefficient as edge information in the approximat subband of lifting wavelet transform, Using edge information and affine model to estimate geometric attack.Then the thesis proposes a self-adapt watermarking algorithm robust to geometric attack in DWT domain. In the algorithm, a watermark was embedded adaptively in low band of DWT domain, according to the conceal quality of Human Visual System (HVS). The proposed watermarking algorithm is not only imperceptible but also robust against geometric distortions, such as translating, rotating, zooming, mirroring, transposing and general image attacks.3. A watermarking algorithm robust to geometric distortion in Contourlet domain. It uses the filiation in different resolutions in Contourlet to put forward“important-tree”, thereby choose the important coefficients to embed and extract the watermarking. Also it uses image normalization, by calculating and comparing the eigenvalue of original image (the watermark has been embedded) and incept image (attacked image) to resume image and extract watermarking information. The results show this method has good geometrical characters resisting to the attack of plane-vertical flip, rotate and scale.更多还原