【作者】 唐坤；

【导师】 葛华勇；

【作者基本信息】 东华大学 ， 信息与通信工程， 2015， 硕士

【摘要】 随着计算机网络和信息技术的快速发展，信息安全受到了各国的重视，用于保密通信的隐写技术成为了研究的热点。隐写技术作为信息隐藏技术主要分支,通过将秘密信息隐藏到公开的数字媒体中,来实现隐秘信息的传递,因而具有更强的隐蔽性。因此，隐写技术在保障个人、商业、乃至国家的信息安全，都有着非常重要的研究意义和实用价值。基于图像的信息隐藏技术是其重要的研究方向之一，由于JPEG图像以其传输容量大、高压缩比，被广泛应用于传输网络中，以该格式图像为载体的隐写算法发展迅速。本文在前人的基础上，以JPEG格式图像为信息隐藏的载体，并且充分利用人类视觉特性，提出了改进的隐写算法，该算法对嵌入容量和安全性都有所提高。同时对YASS算法进行了详细描述，针对YASS隐写算法中存在的缺陷，提出了一些相关改进，主要的工作如下：第一，根据亮度特性和纹理特性，并结合图像分块，提出一种基于视觉特性的图像隐写算法。该方法利用了人眼视觉系统特性，即亮度和纹理,对每个分块利用该特性，将图像划分为强嵌入区、次嵌入区、弱嵌入区，分别嵌入不同比特的秘密信息；在秘密信息嵌入时,利用噪声可见性函数和像素高位平面加权和，计算每一个分块的亮度,接着再对块中像素进行纹理分类，分类后利用模函数算法将隐藏数据非均匀地嵌入在图像亮度分量的DCT低、中频量化系数中，有效地减少了载密图像与原始图像的视觉差异。实验结果表明，利用人类视觉特性改进的隐写算法，可以有效提高秘密信息嵌入容量、抗检测性能，以及保持较高的峰值信噪比。第二，考虑到YASS隐写算法对抗隐写分析能力的不足，本文提出一种YASS改进算法。该算法提出一种最小化嵌入失真的计算方法,对嵌入块修改后的量化DCT系数进行失真分析，选择失真影响最小的交流DCT系数进行秘密信息嵌入，取代传统方法直观选择前19交流DCT系数。实验结果证明，依据嵌入失真由小到大的顺序选择嵌入信道,在抵抗隐写分析和视觉质量方面都优于原始隐写方法。更多还原

【Abstract】 With the rapid development of computer network and information technology, informationsecurity has received the attention of the countries, and the steganography for securecommunications has become the hot spot. Steganography technique is the main branches ofinformation hiding technology, realizing the hidden information by hiding secret information tothe public in the digital media. Therefore, the steganography has very important researchsignificance and practical value in the protection of information security with personal, business,and even the national. JPEG images are widely used in the transmission network, steganographyalgorithm is developing rapidly based on the format of the image. The steganography method hashuge potential for development based on the human visual characteristics, with its hiddenperformance is good and the characteristics of large embedded information capacity. YASS is anovel steganography algorithm based on JPEG image, which can effectively resist the mainstreamof steganography analysis method.In the paper, I put forward the improvement of the steganalysis algorithm, based on the JPEGimages make full use of human visual characteristics as the carrier of hiding information. Theexperimental results show that the algorithm was improved in the embedding capacity andsecurity. At the same time, YASS algorithm was described in detail, for the defects existing in theYASS steganalysis algorithm, I put forward some related improvement. The main work asfollows:First, according to the luminance feature and texture features, connecting with the imageblock, put forward a kind of image steganography algorithm based on visual features. This methodtakes advantage of the human visual system characteristics, namely the luminance feature andtexture features, image is divided into three areas: strong embedding area, time embedded area andweak embedded area using this feature for each partition, and different bits of the confidentialinformation is embedded in different regions; When the secret information embedding, use thenoise visibility function and planar weighted pixels high to calculate each block of brightness and texture features，determining the area of block, hidden data is non-uniformly embedded in theimage luminance component DCT of low frequency and medium frequency quantizationcoefficients, effectively reducing the visual difference between stego-image and the original image.Finally, the results also confirmed that algorithm can improve the capacity of secret information,the performance of resistance testing, and maintain a higher peak signal to noise ratio.Second, considering that the YASS algorithm against steganography analysis ability isinsufficient, this paper proposes a new improved algorithm YASS. Algorithm is proposed tominimize the embedding distortion calculation method, distortion analysis of the quantized DCTcoefficients embedded block modified, the distortion of minimal impact of AC DCT coefficientsare chose to embed secret information, and replace the traditional method directly by choosing thetop19in the AC DCT coefficients. The experimental results show that, according to the orderfrom small to large, select the embedding distortion embedding channel, the improved algorithmis better than the original steganography method in both security performance and visual quality.更多还原