【作者】 张奎；

【导师】 廖述剑；

【作者基本信息】 太原理工大学 ， 通信与信息系统， 2010， 硕士

【摘要】 随着计算机网络技术的飞速发展、公众版权保护意识的增强,数字信息的安全问题越来越重要,数字水印技术应运而生,在数字产品版权保护领域得到越来越多的关注。本文主要研究了空间域和DCT域的数字图像水印技术。空间域技术是通过修改像素的亮度来实现水印的嵌入,具有简单易行,时效性好的特点。离散余弦变换(DCT)是数字信号处理技术中最常用的线性变换之一,该变换具有良好的能量压缩以及去相关的能力,因此将离散余弦变换应用到数字水印处理的技术中逐渐受到重视。通过研究,本文提出了两种基于空域和一种基于DCT域的数字图像水印算法。1.提出了两种基于空域的数字水印算法。利用了将水印能量均匀分布在载体图像的能量上可以提高算法安全性的特性,通过有规律地改变图像像素的数值,实现水印信息的均匀、平滑地嵌入。在嵌入水印前对水印数据进行Arnold变换和混沌加密,增强了水印的抗攻击性能。虽然这是一种明水印算法,提取水印时,需要读取原始图像,但是由于使用了多个密钥,算法的安全性得到了提高。基于量化嵌入可以实现盲提取以及块均值具有良好的承受攻击的能力的特性,根据各个子块的纹理复杂度选择不同的量化步长,保证了水印的不可感知性。计算图像子块的均值,对每个子块的均值进行量化,每个图像子块的整体数据的改变代表一个水印信息。由于使用了量化的方法,提取水印时不需要原始载体图像的参与。2.提出了一种基于DCT域的数字水印算法。该算法利用HVS的空间频率特性,选择合适的嵌入区域嵌入水印,并且对于具有不同纹理特性的区域采用不同的嵌入强度,保证了水印的不可感知性。由于经过各种图像处理后,嵌入位置周围参数的变化可能改变图像子块的原有特性,因此计算嵌入位置周围的DCT系数的均值与水印信息之和嵌入相应的位置,增强了水印的鲁棒性。实验结果表明,本文提出的算法均满足数字图像水印技术不可感知性和鲁棒性的要求。仿真实验中,以有意义的二值图像作为水印信号,结果直观可信。同时,基于均值量化的空域水印算法和基于DCT域的水印算法是盲水印算法,具有一定的实用性。更多还原

【Abstract】 With the rapid development of computer network technology and the enhancement of public consciousness of copyright protection, the security of digital information becomes very important. The digital watermarking technology emerges as the times require. As an effective method for copyright protection, the digital watermarking technology has acquired more and more attention.The main task in this paper is focused on digital image watermarking that based on spatial domain and DCT domain. The principle of digital watermarking based on spatial domain algorithms is that, the watermarker is embedded by modifying the pixel, so it is simple and easy. Discrete cosine transform is one of the most common linear transformations in digital signal processing technology. This transformation has good energy compression and the ability to remove correlation. So it is getting more and more attention in watermarking technology. Through the research, two kinds of digital watermarking algorithms based on spatial domain and one based on DCT domain are proposed.1. Two kinds of digital watermarking algorithms based on spatial domain are brought forward in this paper.(1)When the watermark energy is distributed uniformly in the energy of the carrier image, the safety of the algorithm will be improved. This characteristic is fully utilized in the first algorithm. The watermark is embedded evenly and smoothly by changing the pixels of the image sub-block regularly. Adding a series of pre-processing including Arnold transform、chaos encryption to watermark data before embedding enhanced the anti-attack performance of watermark. The raw carrier image is needed when extracting the watermark, so it is not a blind watermarking algorithm, but by using multiple keys, the algorithm becomes more safety.(2)The mean value of sub-block has good ability to withstand the attack, based on that, the second algorithm is proposed. According to the various sub-block texture complexities, quantization step length is selected, so that the requirements of imperceptibility are satisfied. Then, the mean value of the sub-block is calculated and quantified. Each watermark data is represented by the change of the overall data of the sub-block. By using the method of quantization, the blind image watermarking extraction is realized.2. One kind of digital watermarking algorithm based on frequency domain is presented. Using the spatial frequency characteristics of HVS, appropriate areas are selected for watermark embedding. And each image sub-block which has different texture characteristic is embedded by different strength.Because of that after several kinds of image processing, the parameters around the embedding positions may be changed; the mean value of the DCT coefficients around it is calculated. The values on the embedding positions are replaced by the sum of the mean value and the value of the watermark, so the robustness of the watermark is enhanced.In the simulation experiments, gray image is used as carrier image, and a meaningful binary picture used as watermark signal, the result is intuitionistic and believable. Experimental results demonstrate that the three algorithms can satisfy the requirements of imperceptibility and robustness. And the last two algorithms are more practical, because they do not need the original carrier data when extracting the watermark.更多还原