【作者】 杨恒伏；

【导师】 孙星明；

【作者基本信息】 湖南大学 ， 计算机应用技术， 2009， 博士

【摘要】 自从20世纪90年代以来,随着互联网和多媒体信息处理技术的飞速发展与快速普及,经由互联网与计算机进行数字作品的存储、编辑、复制与传播变得非常容易,并越来越广泛。但这使得数字作品的安全通信与真实性认证受到越来越严峻的挑战。近年来,信息隐藏技术作为解决数字图像安全性问题的重要手段得到了广泛关注,成为近年来信息安全领域的一个研究热点。一个良好的信息隐藏系统应具有鲁棒性、透明性与安全性等基本属性。为获得鲁棒性与透明性的综合平衡,一个有效途径是:结合人类视觉感知特性以开发媒体自适应的信息隐藏算法。本文在充分考虑图像自然特性和视觉感知特性的基础上,研究图像自适应的数据隐藏与脆弱水印技术,主要研究内容与成果如下:1.提出一种高容量图像空域自适应数据隐藏算法。现有许多数据隐藏算法在原图像边缘部分隐藏较多的秘密数据,没能较好地区分图像边缘和纹理,从而导致隐秘图像边缘部分降质较大。为避免秘密数据的嵌入引起图像边缘部分有大的改变,以获得较好的隐秘图像质量,通过利用自适应LSB (Least Significant Bit)替换,提出一种新的图像空域数据隐藏算法。算法利用原图像亮度、边缘、纹理特性建立起一个图像空域视觉感知模型,并应用该视觉模型计算出各像素用于数据隐藏的LSBs位数k,从而使得噪声敏感区像素具有较大的k值,而噪声不敏感区像素则有较小的k值。最后,用一个像素优化调整过程以进一步增强隐秘图像视觉质量。另外,为保证秘密数据嵌入前后各像素用于数据隐藏的LSBs位数k不变,算法利用各像素高r位MSB位来计算k值,而参与数据嵌入的各像素LSBs位不参与k值计算。理论分析与实验结果表明,该视觉感知模型较好地区分了图像纹理与边缘特性,可广泛应用于信息隐藏算法设计。与众多同类算法相比,该提出算法具有更高的隐藏容量与隐秘图像质量。2.提出一种基于视觉感知的图像空域半脆弱水印算法。算法首先利用图像BTC (Block Truncation Coding)域特征经哈希生成依赖于图像内容的水印信号,然后利用自适应LSB替换嵌入水印;最后应用优化像素调整以进一步提高水印透明性。算法从理论上推导出篡改检测与定位自动判别阈值,实现了攻击类型的自动分类,能有效区分恶意攻击和偶然攻击。实验结果表明,算法对偶然攻击有一定的鲁棒性的同时对恶意攻击较为敏感。算法透明性较好,与同类算法相比,显示出较高的定位精度与水印嵌入容量。3.提出一种图像自适应小波域半脆弱水印算法。算法充分考虑了图像视觉感知特性与小波变换的时频局部特性,通过自适应修改图像分块小波域水平细节子带与垂直细节子带系数大小关系以嵌入水印信息。算法透明性较好,人眼感觉不到隐秘图像中水印的存在,嵌入水印与载体图像特征是自适应的。实验结果表明,算法对通常的内容保持操作具有较好的鲁棒性,特别是抗JPEG压缩能力较强。另外算法对恶意攻击较为敏感,并能较好地检测与定位篡改区域。算法能合理区分攻击类型,与同类算法相比,显示出更好的攻击分类性能。4.提出一种基于自适应模运算的无损认证水印技术。通常的基于模256加运算的无损认证水印易造成椒盐噪声,影响隐秘图像视觉质量。为此,充分考虑图像视觉特性,设计一种基于自适应模加运算的无损认证水印算法,以避免椒盐噪声。算法首先结合图像视觉感知特性确定各像素水印嵌入强度,进一步计算出水印嵌入过程中模加运算的自适应除数,然后利用自适应模加运算将水印嵌入到各像素不同的LSBs位中。当隐秘图像未被修改时能无损恢复原图像。实验表明,该算法有效消除了椒盐噪声,极大地增强了隐秘图像视觉效果。与现有无损水印算法相比,该算法显著提高了隐秘图像质量,特别是相对于基于模256加运算的无损认证水印算法,其PSNR值高出约平均9.9864 dB。本文研究将为图像自适应数据隐藏与脆弱水印算法提供新方法与关键技术,研究成果将为互联网图像安全通信与真伪认证提供技术保障,有助于推动数字媒体产业的健康有序发展。更多还原

【Abstract】 With the rapid development of the Internet and multimedia information processing technology since 1990s, the storage, modification, copy and transmission of the digital works via computers and Internet is becoming pretty easier and used more frequently. This causes the secure communication and authentication on digital works to face severe challenges. In recent years, as an effective resolution to the problems on image security, information hiding technique has become a hot research topic in information security field. A well defined information hiding system should have good robustness, transparency and security. To achieve good balance between robustness and transparency, an effective way is to exploit human visual perception, and then develop media adaptive data hiding algorithms. This dissertation focused on image adaptive data hiding and fragile watermarking technique by employing image natural features and visual perception characteristics. The main contributions of this dissertation are as following:1. A high-capacity image adaptive data hiding algorithm in spatial domain.Many existing steganographic methods hide more secret data into edged areas than smooth areas in the host image, which does not differentiate edges from textures and causes serious degradation in actual edge areas. To avoid abrupt changes in image edge areas, as well as to achieve better quality of the stego-image, a novel image data hiding technique by adaptive Least Significant Bits (LSBs) substitution was proposed in this dissertation. Firstly, the scheme developed an image visual perception model by exploiting the brightness, edges, and texture masking of the host image. Secondly, the algorithm used the visual perception model to estimate the number k of LSBs for data hiding. Pixels in the noise non-sensitive regions were embedded by a k-bit LSB substitution with a larger value of k than that of the pixels in noise sensitive regions. Moreover, an optimal pixel adjustment process was used to enhance stego-image visual quality obtained by simple LSB substitution method. To ensure that the adaptive number k of LSBs remained unchanged after pixel modification, the LSBs number was computed by first r most significant bits (MSBs) rather than all the bits of the image pixel value. The theoretical analyses and experiment results showed that the presented perception model could discriminate the edges from textures, and the proposed method achieves higher embedding capacity and better stego-image quality when compared with some existing LSB based data hiding methods.2. A visual perception based semi-fragile watermarking scheme in image spatial domain. In this scheme, the watermark was first generated by the hash of the image feature in Block Truncation Coding domain, and then was embedded into the host image with adaptive Least Significant Bit substitution. Finally, an optimal pixel adjustment was used to further enhance the transparency. The algorithm deduced theoretically the adaptive threshold for automatic tampering detection and location, and could differentiate automatically incidental attacks and malicious attacks. Experiment results showed that the proposed algorithm was sensitive to malicious attacks while robust to incidental attacks, and had good accuracy in detecting and locating tampered blocks. When compared to some existing semi-fragile watermarking methods, the proposed scheme achieved higher embedding capacity and accuracy in tampering location.3. An image adaptive semi-fragile watermarking based on wavelet transform.By taking full advantage of the visual perception characteristics and the local time-frequency characteristics of discrete wavelet transform (DWT), this dissertation presented an image-adaptive semi-fragile watermarking technique operating in DWT domain. The proposed algorithm embedded the watermark by modifying coefficients of the vertical and the horizontal detail sub-bands of image sub-blocks. The algorithm had good transparency, and the watermark was invisible to human eyes and adaptive to the original image characteristics because of the utilization of human visual perception. The experimental results showed that the proposed algorithm was effective and can distinguish adaptively some non-malicious image operations such as compression from malicious ones made on the watermarked image. The embedded watermark was sensitive to malicious attacks while robust to non-malicious attacks, especially to JPEG loss compression. Moreover, the scheme outperformed other semi–fragile watermarking algorithms at the classification of attacks.4. A lossless authentication watermarking based on adaptive arithmetic.Reversible watermarking schemes based on modulo-256 addition may cause annoying salt-and-pepper noise. To avoid the salt-and-pepper noise, a reversible watermarking scheme using visual perception characteristics and adaptive modular arithmetic was proposed. First, the high bits residual image was obtained by extracting r MSBs of the original image, and then the watermark strength and the adaptive divisor of modulo operation in each pixel were computed by employing the high bits residual image features. Finally, the watermark was embedded into different LSBs of original image with adaptive modulo addition. The original image could be lossless recovered if the stego-image had not been altered. Extensive experiments showed that the proposed algorithm eliminated the salt-and-pepper noise effectively, and the visual quality of the stego-image with the proposed algorithm had been dramatically improved over some existing reversible watermarking algorithms. Especially, the stego-image of this algorithm had 9.9864 dB higher PSNR value than that of modulo-256 addition based reversible watermarking schemes, on average.This dissertation provided new methodology and key techniques for image adaptive data hiding and semi-fragile watermarking algorithms. Research achievements in the paper will offer secure guard for secure image communication and image authentication via the Internet, and be help to promote the healthy and orderly development of digital media industry.更多还原