【作者】 金喜子；

【导师】 王钲璇；

【作者基本信息】 吉林大学 ， 计算机应用技术， 2011， 博士

【摘要】 随着Internet的迅速发展,多媒体信息交流变得越来越方便,但由于数字媒体很容易被非法复制,因此信息安全保护问题日益突出,如何有效地防止数据被非法复制及鉴别多媒体产品的知识产权,已成为一个亟需解决的问题。数字水印技术在这种背景下应运而生,它在信息安全和版权保护方面起着重要的作用。数字水印技术是在多媒体载体中隐藏一些标志性信息,但载体本身内容的价值和使用却不受到任何影响,且人的感知系统也不能察觉这些信息,以便在版权纠纷中用来证明版权归属(侵权行为)以及认证数字产品内容的完整性。数字水印技术作为一种数字媒体版权保护、数据认证、数据隐藏等方面的有效办法,得到了国内外学者的广泛关注。本文主要研究了数字水印技术在数字图像方面的应用,首先介绍了数字水印技术的研究背景和发展现状,进而对数字水印的原理、分类、典型算法、应用领域及评价标准做出了简要分析,然后以静止图像为研究对象,通过对DCT域数字水印技术的特点进行深入的研究和探讨,提出了几种基于DCT域的数字水印算法。主要研究内容如下:1.提出一种新的多比特水印嵌入方案,用多个伪随机序列负载更多比特的水印信息。本方案将载体灰度图像做8×8的分块DCT变换,变换后每一块中的不同系数代表载体不同频率分量。将每一小块中(4,4)频率位置的DCT系数提出构成一个数据块,并将这个数据块以一定的方式置乱为一个一维向量X,长度设为N。然后根据密钥伪随机地产生N个长度为N的正交序列,并选择其中的M个序列(L比特,L>M)以一定强度迭加到一维向量X上,选择的方法由嵌入的水印信息来决定。由于这种方案是利用载体数据与伪随机序列相关系数的大小关系负载水印信息的,因此嵌入的比特数L大于迭加的伪随机序列数M,提高了水印性能,特别是当N较大时,L更是远大于M。本方案将水印信息嵌入在中频分量上,兼顾了稳健性和隐蔽性。本方案还对不同参数进行了隐蔽性和稳健性的测试,实验结果表明稳健性主要与系统参数A有关, M的变化对稳健性影响不大;而嵌入量L则取决于参数M;隐蔽性则与参数A和M都有密切关系,实际应用中,可以根据不同情况调整这两个系统参数。2.JPEG是一种常见的图像格式,在JPEG图像中进行准确的篡改定位具有重要意义。JPEG编码过程中,首先将图像分成8×8的小块,在每个小块内进行DCT变换,然后将不同位置的DCT系数按照不同的步长量化为整数。为了减少水印信息量,本文方法并不是将每个小块的Hash直接嵌入载体图像中,而是先求出每个小块主要内容的Hash,然后将Hash比特重新分组并作模2和,并将这些模2和的结果作为真正的水印信息。水印比特数与JPEG图像的小块数一致,因此在每个小块中仅需嵌入1比特水印,保证了良好的隐蔽性。认证时根据待认证图像内容计算水印信息,并与从待认证图像中提取的水印信息进行比较,根据整体匹配情况估计篡改率,再根据每个小块对应的水印信息的被破坏程度判别该小块是否曾被篡改。当篡改区域小于1%时,本文方法可以准确定位所有篡改小块。如果对含水印图像进行不同质量因子或不同量化矩阵的重新编码、或在含水印图像中迭加较强的噪声,这相当于对含水印图像进行大范围的篡改操作,因此不能定位篡改位置。理论分析准确的篡改定位能力极限并设计可准确定位更大篡改区域的脆弱水印方法将是下一步研究内容。3.提出一种基于DCT变换和DWT变换的抗JPEG压缩的半脆弱数字水印方案。本方案首先对原始图像进行量化和反量化处理后得到载体图像,并利用载体图像经DWT变换得到的LL3子带图作为原始水印图像,且只选取水印图像的高四位作为水印信息。然后对载体图像进行8×8的分块DCT变换,并充分融合水印嵌入和JPEG压缩过程,根据密钥选择在载体图像DCT系数的低频分量上嵌入水印信息。提取水印时,只需根据相应位置DCT系数的奇偶性即可提取出水印的高四位信息。由于在水印检测时,不需要使用原始图像就可实现水印的检测,因而是一种完全盲数字水印算法。实验结果表明,当质量因子为50时,提取出的水印图像的质量非常清晰,与原始水印图像几乎相同,并且当质量因子为40时也能成功提取出水印图像,说明本方案对JPEG压缩具有良好的鲁棒性。另外,本方案对噪声攻击和裁剪攻击也具有较好的鲁棒性,并能进行一定程度的篡改定位。更多还原

【Abstract】 With the rapid development of Internet, the communication of multimedia becomes more and more convenient. But because of the digital media is very easy to be illegal copied, information security becomes increasingly outstanding. How to efficiently prevent the illegal copying of data and identify intellectual property rights of multimedia product is a pressing problem. In this background, the digital watermark technology emerged at the proper moment, which plays an important role in information security and rights protection.Digital Watermarking technology will hide the symbolic information in multimedia carrier, which should not influence the quality of the multimedia carrier obviously and should not be perceived by people’s perceptual system, so that it can prove the ownership of copyright (infringing act) and integrity the authentication of digital products in copyright disputes. Digital watermarking technology as an effectual means about digital media copyright protection, data authentication, data hiding and so on has gotten the attention of scholars both at home and abroad.This paper mainly researches the digital watermark technology applied in digital image. At first, the research background and current development status of digital watermark are introduced. Furthermore, it makes a brief analysis about principle, classification, typical algorithms and standards of digital watermark. This paper is a study of static images, with a deep research about the several digital watermarking algorithms based on DCT are proposed in the paper. And the main research contents are as follows:1.This paper proposes a multi-bits watermarking scheme in which carrying more information of watermark using multi-pseudo-random sequences. In this scheme, the 8×8 bock based DCT is performed to the host gray level image and the coefficients of a block in the transformed domain represent different frequency component of the host image. Select the DCT coefficient on position (4,4) of every block to form a dataset and permute it to a vector X with length N. Then generate orthogonal sequences with length N randomly according to a private key and select M sequences to add to vector X with certain intensity. The way of selecting is determined by the watermark. This scheme uses the comparison of correlations between the host data and pseudo-random sequences to carry the watermark, therefore providing a payload significantly more than the number of added sequences and enhancing the performance of watermarking. Especially when N is large, L is much larger than M. The watermark is embedded into the middle frequency component of DCT domain, taking into account the robustness and imperceptibility. This scheme have tested the robustness and imperceptibility with different parameters and the results show that the robustness is mainly related to system parameter A and M have little effect to it; the embedding capacity is determined by M; the imperceptibility is related A and M. In the practical application, we can adjust these two parameters according to the actual requirement2.JPEG is a common image format and it is significance to locate the tampered area accurately in the JPEG image. In the process of JPEG encoding, the host image is partitioned to blocks sized 8×8 and DCT is performed to them, then quantify the DCT coefficients to integers with different quantify steps. In order to reduce the amount of watermark information, we do not embed the Hash values of every block to the host image directly. Instead, we calculate the Hash values of the block’s principal content first, then re-group the Hash values and sum them with modulus 2, and finally we get the genuine watermark. The bits of watermark are consistent to the blocks of JPEG image, so one bit is embedded into one block and good imperceptibility is guaranteed. When authentication is performed, we comprise the calculated and the extracted watermark of the image to be authenticated and estimate the tamper rate by the overall match. Then we determine if the block is tampered or not according to the damaged degree of the watermark information of the block. This scheme can locate all tampered blocks accurately when the tamper rate is less than 1%. We can’t locate the tampered area when the stego-image is re-encoded with different quantization factor or quantization matrix, or added by high level noise because these operations are equivalent to tamper the image with large range. Theoretical analysis to the ability of tamper location and design of a fragile watermarking scheme that can locate more extensive tampered area accurately are the future works.3.This paper proposes a semi-fragile watermarking scheme based on DCT and DWT that can resist JPEG compression. In this scheme, we first get the host image by quantization and inverse quantization operations to the original image. Then perform DWT to it and just select the high four bits of the LL3 sub-band as a watermark. In the next step, we perform 8×8 block based DCT to the host image and embed the watermark into the low frequency component selected by private key of DCT domain integrating the embedding and JPEG encoding process. On the extracting end, we extract the high four bits of watermark according the parity of the DCT coefficients. This scheme can detect the watermark without the original image which is a completely blind watermarking algorithm. The experimental results show that the extracted watermark is clear and is very close to the original watermark when the compression quality factor is 50, and the watermark can be extracted successfully also when the quality factor is 40. The result illustrates that this scheme is robust to JPEG compression. This scheme is also robust to noising and cropping attack, and can locate the tampered area to a certain extent.更多还原