【作者】 刘劲；

【导师】 康志伟；

【作者基本信息】 湖南大学 ， 信号与信息处理， 2006， 硕士

【摘要】 密写在军事、情报、国家安全方面具有重要作用。密写技术有两个基本技术要求:隐蔽性和足够的数据量。基于人眼视觉特性的密写方法具有很好的视觉隐蔽性,在众多密写方法中嵌入数据量是最大的。因此,基于人眼视觉特性的密写在密写研究领域中占据着重要地位。但由于近期来,针对基于人眼视觉特性密写的密写分析方法不断地被提出。传统的基于人眼视觉特性的密写的安全性降低了。因此迫切需要设计出高度安全的新密写方法。在这种情况下,本文尝试将小波变换和自组织竞争神经网络应用于图像密写,提出性能更好的新密写方法。本文主要工作如下:(1)深入研究人眼视觉系统、小波变换理论、自组织竞争神经网络的基本理论以及传统基于视觉特性的密写及密写分析。(2)根据人眼对纹理区域噪声不敏感的特性,提出了一种基于小波变换和模算子的密写。在该方法中,较多秘密信息嵌入到强纹理区域。通过试验可以证明与BPCS、PVD和双边匹配相比,该方法具仃优越性。(3)指出了传统小波域密写方法会导致小波直方图的异常,因而存在安全漏洞;为了提高密写的安全性并利用人眼视觉特性,提出了一种基于HVS的抗统计分析的小波域密写。通过试验证明该方法不仅保持了小波直方图特性,而且具有较好的视觉隐蔽性。(4)根据对比度和纹理敏感性,并利用自组织竞争神经网络的分类能力,提出了一种基于自组织竞争神经网络和HVS的密写。实验结果表明,与四边匹密写相比,该方法有更大的嵌入量并保持了良好的载密图像质量。本文使用MATLAB作为仿真实验平台对上述研究进行了验证和分析。更多还原

【Abstract】 Steganography takes an important position in military affairs, information and security of country. A steganographic technique should generally possess two important properties: good imperceptibility and a sufficient payload. The steganography based on HVS has good visual imperceptibility, and its payload is the most in all steganographic methods. Therefore, the steganography based on HVS occupies an important position in the research of steganography. However, recently, the steganalysises of steganography based on HVS were continually proposed. The security of traditional steganography based on HVS is lowered. So designing new steganography with high security is exigent. Under this situation, we try to apply wavelets and self-organizing competitive NNs to image steganography, proposing new steganographic methods with better performance. The main important work of this paper:1. Deeply studied human visual system, wavelet analysis theory, the basic theory of self-organizing competitive NNS , traditional steganography based on HVS and steganalysis.2. In accordance with the visual property that human eyes are less sensitive to strong texture, a novel steganographic method based on wavelet and modulus function was presented. In this method, more secret information is embedded into blocks with strong texture. From the experimental result, the proposed method has a better performance than BPCS, PVD and Two-sided side match.3. Point out that traditional steganography on wavelet domain result in abnormity of histogram for wavelet coefficients, so a loophole exists; in order to improve the security and exploit the HVS, a steganography on wavelet domain based on HVS against statistical analysis is proposed. From the experimental results, the proposed method effectively keeps the identity of histogram for wavelet coefficients and maintains a good visual quality of stego-image.In view of contrast and texture sensitivity, and exploiting the classified ability of self-organizing competitive NNS, a steganographic method based on self-organizing competitive NNs and HVS is proposed. Simulation experiments show that the proposed method has much greater hiding capacity than four-sided side match, maintaining a good visual quality of stego-image.We used MATLAB as the simulation platform to verify the research mentioned above.更多还原