【作者】 苗锡奎；

【导师】 孙劲光；

【作者基本信息】 辽宁工程技术大学 ， 计算机软件与理论， 2009， 硕士

【摘要】 数字水印技术是近十几年来提出的一种有效的数字产品版权保护技术。但目前每一种水印算法是不可能,也根本做不到抵抗所有的攻击。研究的目标往往是针对某一类的攻击而设计算法,并尽量考虑兼容其它种类的攻击。同样水印产品应用的领域往往也各不相同；不同的水印应用环境往往青睐于具有抵抗某一类攻击的算法。因此研究针对某一类攻击的水印算法也具有极其重要的意义。本文重点针对剪切、行列去除、旋转三种类型的几何攻击,分别设计了三种算法。同时,设计算法时还考虑了其它的攻击,应用相应的理论使算法尽可能的抵抗更多的攻击,进而使算法的应用领域得到扩展。1.提出了一种双重变换域和水印划分策略的水印算法。算法首先将水印加密置乱,然后将水印和载体图像分为两部分,将一部分水印自适应嵌入到DCT域视觉容量大的DC系数上,另一部分自适应嵌入到DWT域的中频子带中,实现了嵌入强度与区域特征的自适应。有效的解决了水印的容量问题,并提高了水印的鲁棒性和不可见性；实验表明,算法对JEPG压缩、几何剪切等攻击均达到了预期的指标。2.以图像归一化技术能够良好的抵抗全局几何攻击和分形局部化水印为基础,提出了一种基于图像归一化的分形水印算法。算法首先利用基于矩的归一化技术将载体图像映射到具有几何不变性的空间内；然后再结合不变质心理论在归一化图像中提取重要区域；最后利用分形,通过调整DCT中频系数,将水印自适应地嵌入到重要区域的自相似块中。实验表明,算法不仅具有较好的透明性,而且对常规信号处理攻击和行列去除等几何攻击均具有较好的鲁棒性。3.结合图像归一化、分形理论及伪Zernike矩,提出了一种可以抵抗不同角度的旋转攻击和组合攻击的水印算法。针对归一化图像大小不一致,利用伪Zernike矩良好的旋转不变性来弥补其不足。同时结合分形局部化水印来抵抗剪切攻击,通过对伪Zernike矩的二次测试,将水印嵌入到最鲁棒的自相似块的伪Zernike矩中。实验表明,算法不仅具有较好的透明性,而且对常规信号处理和不同角度的旋转等攻击均具有较好的鲁棒性。更多还原

【Abstract】 Digital watermarking is a potent method for multimedia content protection such as protection of copyright, and becomes significant disquisitive field in the information hiding.But,at present,not every watermarking algorithm is rubust to all the attacks.The aim to designing watermarking algorithm is to against a particular type of attacks,and try to take other types of attacks into account. As the watermarking product application areas are often varying,and different watermarking application environments tend to a certain algorithm against a type of attacks.Therefore,the research for a certain type of attacks watermarking algorithm is also of critical importance.This article is focusing on cutting,removal of rows and columns and rotatation geometric attack, three kinds of algorithms were designed.1.A meaningful digital watermarking algorithm is proposed based on combination of DCT and DWT.Firstly,making watermarking chaos using arnold transform;Secondly, cutting the choas into two pieces. Second, embeding one piece into DCT’s DC coefficie-nt coefficient by adaptive quantization,and embeding the other into DWT sub-band.2. A new fractal watermarking algorithm robust to removal is proposed based on image normalization and fractal.Firstly,the geometrically invariant space is built by using image normalization.Then,a significant region is gained from the normalized image by using invariant centroid principle.Finally, the watermark is embedded into the self-similar blocks DCT transformed of the significant region by fractal.3. A new image watermarking algorithm robust to different angle rotation is proposed.Firstly,significant region is extracted by using normalization and invariant centroid principle.Then,significant region is divided into two groups:self-similarity and non-self-similarity blocks by fractal coding and pre-set threshold,the robustest pseudo-Zernike moments is picked up among the pseudo-Zernike moments of the self-similarity blocks.Finally,the watermark is embedded in the robustest pseudo-Zernike moments by quantization.更多还原