【作者】 郑建勇；

【导师】 陈纯； 陆系群；

【作者基本信息】 浙江大学 ， 计算机应用技术， 2006， 硕士

【摘要】 随着数字技术和因特网的发展,各种形式的多媒体数字作品被通过网络进行传输,然而数字作品的便利性和不安全性是并存的,它可以低成本、高速度地被复制和传播,这样就为创造者和使用者都提供了很大的便利,但这些特性也容易被盗版者所利用,因而采取多种手段对数字作品进行保护、对侵权者进行惩罚已经成为十分迫切的工作。除了与传统作品版权保护相类似的法律和管理手段外,还应该针对数字作品本身的特点为其提供技术上的保护。数字水印技术的研究就是在这种应用要求下逐步发展起来成为国际上研究热门的课题。 本文通过对数字水印技术特点和发展现状的分析,在研究国内外提出的数字水印算法的基础上,对灰度图像和彩色图像数字水印技术的一些典型算法进行了分析和讨论。 一是对灰度图像数字水印空域算法和变换域算法进行了研究和分析。首先,实现了LSB算法,运用常用信号处理对嵌入水印的图像进行了攻击,发现本算法对按比例缩放具有很强的稳健性;对图像剪裁,提取的水印不完整;对其它信号处理抵抗力很弱。其次,对DCT域水印算法进行了改进,改变将水印信息嵌入低频分量的做法,将水印信息直接嵌入DCT系数的直流分量。实验结果证明,将水印嵌入DC系数,水印对大多数信号处理具有较好的鲁棒性,但这种方法的局限性在于嵌入的水印容量太小会,嵌入水印是拉伸因子取值不宜太大,否则会影响原始图像品质。第三,分析了一种稳健性好的DCT盲水印算法。采用双伪随机序列,将水印信息嵌入DCT变换的中频分量,基于阈值相关判断进行水印提取,不依赖原始图像,更有使用价值。嵌入的水印不易在诸如有损压缩等信号处理中被破坏,也能使隐藏的信息不会对原始图像品质产生明显的影响。 二是对彩色图像数字水印空域算法和变换域算法进行了研究和分析。首先,将水印信号转化成二进制数据码流,直接嵌入原始图像的RGB空间每个字节的最低位。该算法优点是嵌入信息的容量与所选取的掩护图像大小成正比,具有较高的信息隐藏率,且对大多数普通信号处理具有一定的鲁棒性。其次,研究分析了一种基于Fourier变换的彩色图像水印算法,对原始图像做Fourier变换,将调制后的水印信息依次加入某些固定位置的幅值谱上。该算法将原始图像进行分类,对白色像素较多的嵌入水印的强度较大,这样可以使透明性和强壮性达到较好的统一;对同一幅图像嵌入水印和提取水印的参数不同,这样尽可能排除由于图像空域变换造成的能量损失。该算法具有较好的鲁棒性,对剪裁、噪声、压缩、涂改、平移等具有较强的抵抗更多还原

【Abstract】 With the development of digital technology and internet, many kinds of digital multimedia are transported through Internet. Digital multimedia could be replicated and transported rapidly with low costs without permission. However these characteristics of digital multimedia transmitted over the Internet would be easily attacked by the pirates. It is very important to find new ways to protect the digital works against the attackers. Besides the traditional protection laws and administrative methods, it is important to protect the digital content utilizing their instinctive characteristics. So the digital watermarking technique is becoming one of the hot research projects around the world.In this paper we study the main features of the current techniques, and analyze the typical watermarking algorithms for the gray images and the color images.First, the watermarking techniques based on spatial domain and transformation domain for the gray images are studied. The LSB algorithm is implemented and the robustness of this algorithm is tested under several kinds of attacks. This algorithm could resist the zooming attack but it is not robust to clipping and other signal processing. Then an improved algorithm based on DCT transform is proposed. The watermark is embedded into the DC coefficients after the DCT transform not the AC components. The experimental results show that the technique could resist most of signal processing attacks. However, the watermarking payload of this method should be lower, otherwise the quality of the original image would be degraded. We also evaluate an adaptive blind image watermarking algorithm based on the block DCT transform. The watermark is a visible binary image. We use two independent pseudo-random sequences to substitute the ’O’andT in the binary image, and the original spectrum will be spread over a wide area. The modulated signal is embedded in the low-middle frequency bands of the block DCT transform, and the scheme can adaptively adjust the strength of the embedding signal according to the original magnitude and can extract the watermark without the original image. The simulation results demonstrate that the method is robust to most of the image procession operations such as lossy JPEG compression, low pass filtering, median filtering, cropping etc.Second, we analyze the watermarking algorithms based on spatial domain and transformation domain for color images. Firstly, we transform the watermark signal into binary data stream and embedinto the least significant bit of every byte in RGB color space. The capacity of the watermark image scales the size of the original image. This method holds high data hiding rate. It is robust to most of normal signal processing attacks. A valid blind watermarking algorithm is introduced here. The images that have a lot of white pixels usually can produce a lot of overflows after the watermark is embedded. The algorithm embeds the watermarking signal modulated on some fix positions which keep a certain distance from each other. Some decimals are added to the image-signed and different parameters are embedded for different images. The experiments show that this algorithm is transparent, robust, accurate and cost-efficient.Third, this paper studies two watermarking algorithms based on the wavelet domain. One is to embed the watermark sequence into LH and HL coefficients of wavelet transforms. We use two independent pseudo-random sequences to substitute the ’O’andT in the binary image, which has the same effect of the spread spectrum. So this algorithm has a better quality of robustness and a better quality of resisting the damage from clipping, noise, JPEG compression, modification and shifting. Another is to embed a 24-bit color image into the HH coefficient of another 24-bit color image. Color spaces with linear relation to RGB color space with uncorrelated components are found to be most suitable for watermarking application. The YUV color space is more suitable to embed watermark. We first transform the RGB color space to the YUV color space. Then we apply wavelet transformation to the respective YUV components of the original image and watermark image. The YUV components of the watermark image are embedded into the corresponding YUV components of the original image. The future research work is to find the more effective embedding location in the original image.更多还原