【作者】 李鹏；

【导师】 马培军；

【作者基本信息】 哈尔滨工业大学 ， 计算机应用技术， 2012， 博士

【摘要】 秘密图像信息的安全保护是信息安全的一个重要研究领域。图像秘密共享主要解决秘密图像信息在存储和传输时容易丢失和被篡改的问题，提高秘密图像保护的安全性。目前，图像秘密共享方法的研究主要可分为两大类：视觉密码共享和基于多项式的图像秘密共享。前者适用于在秘密图像恢复阶段无计算条件，且对恢复图像视觉质量要求不高的情况；后者适用于在计算机的辅助下，通过计算恢复精确的秘密图像信息。本论文从提高共享效率和视觉质量角度来研究图像秘密共享方法。首先分别针对基于多项式的图像秘密共享方法和视觉密码共享方法的性能提高进行研究，然后研究视觉密码共享方法和基于多项式的图像秘密共享方法的融合问题。主要研究内容包括：（1）针对现有的基于多项式的图像秘密共享隐藏方法中，载密图像存在较大的尺寸扩张和视觉质量降低的问题，提出一种基于最大峰值信噪比估计的图像秘密共享隐藏方法。首先从理论上分析在不同的信息嵌入方式下，载密图像的峰值信噪比估计值；然后给出载密图像具有最大峰值信噪比估计值时影子图像的嵌入算法，对掩盖图像尺寸大小与载密图像的视觉质量进行联合优化，实现载密图像尺寸自适应的图像秘密共享隐藏。载密图像较好的视觉质量可以通过较大尺寸为代价来获得。针对载密图像合法性验证的问题，提出分段验证算法，能够有效的验证载密图像的合法性。（2）针对传统视觉密码共享存在较大的影子图像尺寸扩张，并且容易导致恢复图像纵横比改变的问题，研究纵横比不变的视觉密码共享方法。首先利用图像压缩的方法解决影子图像尺寸扩张的问题。在此基础上，利用秘密图像尺寸调整的方法实现任意大小的尺寸扩张。为了提高恢复图像在细节上的视觉质量，将秘密图像中的边缘像素信息优先保留到恢复图像中。最后分别基于确定型的视觉密码和概率型视觉密码提出纵横比不变视觉密码共享方法。解决了纵横比不变的视觉密码共享在尺寸扩张系数小于1时，恢复图像视觉质量严重降低的问题。（3）针对图像秘密共享方法无法同时实现快速解密和精确解密秘密图像的问题，研究结合视觉密码和多项式的图像秘密共享(Visual Cryptography andPolynomial-combined Image Secret Sharing,简称VCPISS)方法。在恢复过程中，既可选择叠加影子图像解密模糊的预览图像，也可利用计算恢复出精确的秘密图像。首先分析文献方法中存在的两个不足：叠加影子图像的解密图像视觉质量低和计算解密后的图像有损；然后给出无损恢复的改进VCPISS方法。针对叠加影子图像后的恢复图像视觉质量低的问题，提出基于灰度视觉密码的VCPISS方法，利用较小的尺寸扩张为代价获得较大的视觉质量上的改善。为了更好的衡量叠加影子图像得到的恢复图像视觉质量，提出了新的对比度定义。在此定义下，可以有效地评价灰度视觉密码的恢复图像视觉质量。（4）为了实现在不同的门限和解密条件下恢复出不同视觉质量的秘密图像信息，研究双重门限的图像秘密共享方法。分别基于传统视觉密码和灰度视觉密码，提出了两种双重门限图像秘密共享方法。在基于传统视觉密码的双重门限图像秘密共享方法中，通过选取不同类型的共享矩阵在视觉密码的影子图像中嵌入秘密图像信息。共享矩阵按两两不同划分成不同的类型，从而提高信息嵌入的容量，进而在视觉质量不变的前提下得到较小尺寸的影子图像。在基于灰度视觉密码的双重门限图像秘密共享方法中，除了利用不同共享矩阵类型来嵌入信息外，直接在每个影子像素中嵌入灰度值信息，从而大大提高了影子图像中嵌入信息的容量，以较小的视觉质量降低为代价获得更小的影子图像尺寸扩张。综上所述，本文对视觉密码和基于多项式的图像秘密共享方法，以及两种技术的融合方法进行了深入研究，在提高图像秘密共享方法的共享效率和视觉质量等方面取得了很好的效果。更多还原

【Abstract】 The secure protection of secret image is an important research area ofinformation security. Image secret sharing (ISS) mainly resolves the problems oflosing and modifying of the secret image information in storage and transmission,and improves the security of the secret image’s protection. Currently, the research ofimage secret sharing techniques can be divided into two categories: visualcryptography sharing (VCS) and polynomial-based ISS (PISS). VCS is suitable forthe situation where no computation resources are available, and the perfectrevealing of the secret image is not required. PISS needs computer to reveal theperfect secret image. This dissertation mainly addresses the researches of ISSmethods by improving efficiency and visual quality. It first improves theperformances of PISS and VCS, respectively. Then the combining problem of VCSand PISS is researched. All research contents are as follows.(1) Aiming at large size and low visual quality of the stego image inpolynomial-based image secret sharing and hiding (PISSH) method, the dissertationproposes a PISSH method based on the largest PSNR estimation. The PSNR value ofstego image is estimated by different information hiding methods, theoretically.Then a hiding method of shadow image is proposed based on the largest PSNRestimation. It can jointly optimize the size and visual quality of stego image. Thesize of stego image can be chosen self-adaptively. The better visual quality of thestego images can be achieved with the cost of larger size. According to the problemof verifying the validity of the stego image, a piecewise authentication algorithm isproposed, which can verify the validity of the stego image effectively.(2) Aiming at the large size expansion of shadow image in traditional visualcryptography sharing, and distortion of the revealed image, the dissertationresearches the method of aspect ratio invariant visual cryptography sharing(ARIVCS). It first realizes shadow image without size expansion by compressingthe secret image before sharing. Then optional size expansion of shadow image isachieved using the resizing process of the secret image. In order to improving thevisual quality of the revealed image in details, the edge secret pixels have priorities to be retained in the revealed image. Finally, the dissertation proposes two ARIVCSmethods based on deterministic VCS and probabilistic VCS, respectively. It solvesthe problem of degraded visual quality in VCS when the size expansion is smallerthan1.(3) A single image secret sharing method cannot simultaneously realize fast andperfect revealing the secret image. The dissertation researches the method of visualcryptography and polynomial-combined image secret sharing (VCPISS). In therevealing process, it can reveal a vague previewed image by stacking shadowimages, also a perfect image by computation. Two shortcomings of a literatureVCPISS method are analyzed: low visual quality of the previewed image bystacking shadow images, and lossy reconstructed secret image by computation. Animproved VCPISS method of lossless reconstruction is then proposed. In order toimprove the visual quality of the previewed image by stacking shadow image, aVCPISS method is proposed based on gray VCS (GVCS). The visual quality of thepreviewed image can be largely improved with a small cost of size expansion. Toevaluate the visual quality of the previewed image more reasonable, a newdefinition of the contrast is proposed. Under the new definition, the visual quality ofthe revealed images in GVCS can be compared effectively.(4) Aiming at revealing secret image with different visual quality in differentthresholds and decoding conditions, the dissertation researches the method ofdouble-threshold image secret sharing (DTISS), and proposes two DTISS methodsbased on VCS and GVCS, respectively. In the VCS-based DTISS method, secretimage information is embedded into shadow images of VCS by choosing differenttypes of share matrixes. All pair-wise different share matrixes are divided intodifferent types. Therefore, larger information can be embedded into the shadowimage, and get smaller size of shadow image without degrading visual quality. Inthe GVCS-based DTISS method, except embedding information by choosingdifferent types of share matrixes, each shadow pixel embeds gray-valuedinformation. Hence it greatly improves the storage of the embedded information inshadows. Therefore, the GVCS-based DTISS method can greatly reduce the sizeexpansion of shadow image with a small cost of visual quality.In conclusion, this dissertation has researched the methods of visualcryptography and polynomial-based image secret sharing, and the combination of these two techniques. It achieves good performances in two key problems: thesharing efficiency and the visual quality of the revealed image.更多还原