【作者】 边杏宾；

【导师】 朱清新；

【作者基本信息】 电子科技大学 ， 计算机应用技术， 2009， 博士

【摘要】 随着互联网的普及和数字技术的广泛应用，数字产品变得极大丰富并且传播便捷，同时版权保护问题也日益突出。传统的信息安全技术在数字产品版权保护方面存在的不足，促成了数字水印技术的发展。数字水印技术将数字水印隐藏于数字媒体中，以便在版权纠纷中为版权所有者提供版权证明。鲁棒性和隐蔽性是版权保护数字水印的一般特点，盲提取能力能方便数字水印技术的应用。本文针对数字水印技术的特点及其应用作了以下探索和研究：1．提出了基于模板匹配的鲁棒性数字水印方案。该方案依照模板，将数字水印嵌入到图像的小波变换域，实现了数字水印提取过程的自同步和盲提取。在提取数字水印前，通过对宿主图像的锐化处理，能够显著改善数字水印的提取精度，并且能够提高数字水印抗图像处理和有损压缩等攻击的能力。2．提出了图像归一化特征区域数字水印方案。该方案消除了基于不变矩的图像归一化(MIBIN)在数字水印技术应用中的问题。在数字水印技术中，MIBIN被用于应对几何变换攻击，但诸如嵌入的数字水印、灰度直方图调整、图像的少量剪切等因素影响归一化图像的稳定性，从而影响数字水印的提取，降低数字水印的鲁棒性。从MIBIN的原理出发分析了上述问题，并在该方案中给出了解决方法。3．提出了基于支持向量机(SVM)的数字水印方案。借助SVM良好的学习能力和推广能力，该方案提高了数字水印的鲁棒性，并且实现了不依赖任何辅助信息的数字水印盲提取。该方案利用了数字水印的提取与模式识别的内在相似性，设计了适合于SVM应用的数字水印嵌入和提取方法，分析了SVM的核函数及其参数选取在水印提取效果方面的表现。4．提出了对视频具有双重保护的数字水印方案。从视频编码原理出发，分析了在视频中嵌入数字水印后，数字水印和视频之间的相互影响，由此得出了视频重建误差(VRE)的产生和消除机制。该方案利用这种机制提高了数字水印的鲁棒性，并且对视频提供双重保护，既能够由数字水印提供版权证明，同时又利用VRE直接保护视频内容。5．提出了以存储为中心的网上交易与认证结构，以及该结构下的数字水印协议。解决了数字水印协议中数字产品信息和交易信息的存储问题，保障了数字产品和交易记录存储的一致性，便于集中管理信息，克服了Buyer-Seller水印协议中的若干安全问题，而且使子协议之间的划分更清晰。6．提出了数字水印空间概念，通过改进离散空间静态目标的最优搜索模型和求解方法，建立了数字水印空间中多副本水印的最优提取序列。多副本数字水印方案能够全面提高数字水印的鲁棒性，但是受到攻击后不能确定哪个数字水印副本存在，如果采用穷尽搜索的方式会花费大量时间，而最优提取序列能使提取到数字水印的期望时间最小。更多还原

【Abstract】 With the rapid development of Internet and digital technology,digital productsbecome more and more popular,mainly because they can be transmitted conveniently,but the problem of copyright protection is also more and more serious.Theinapplicability of traditional information security technology for copyrights problempromotes the development of digital watermarking.Digital watermarking technologyembeds digital watermark in digital media that provides copyright proof for the ownerof digital media.Robustness and invisibility are two important features of digitalwatermark.The blind detection property can facilitate application of digitalwatermarking.This dissertation thoroughly studied the watermarking technology and theapplications.The main contributions are listed below:1.A robust digital watermarking scheme based on template matching is proposed.The digital watermark is embedded in the wavelet transform domain according to atemplate.This scheme realizes blind watermark extraction without any extrainformation.By sharpening the watermarked image before extraction,the precision ofthe extracted watermark is enhanced,thus the robustness of the watermark is improvedagainst image processing and lossy compression etc..2.A new scheme for embedding watermark in normalized feature area ispresented,that eliminates some problems in MIBIN (Moment Invariant Based ImageNormalization).MIBIN has been used in watermarking schemes to resist geometrictransform attacks,however,the stability of the normalized image and watermarkextraction can be affected by the embedded watermark,gray level histogram adjustment,and content removal etc.Thus the robustness of the digital watermark is damaged.FromMIBIN theory,some solutions to these problems are given in this scheme.3.A watermarking scheme based on SVM (Support Vector Machine) is presented,the robustness of the scheme comes from the good learning and generalization ability ofSVM,the watermark can be extracted without any auxiliary information.Using thesimilarity of digital watermark extraction and pattern recognition,this scheme gives methods for watermark embedding and extraction,which are suitable to SVMapplication,by analyzing the influence of kernel functions and their parameters onwatermark extraction.4.A digital video watermarking scheme with dual protection is presented,whichcan provide copyright proof by the watermark,and protect the video content directly bythe VRE (Video Reconstruction Error).From the video coding procedure,theinteraction between the watermark and the video is analyzed,and the mechanism ofproducing and elimination of VRE is obtained.This mechanism is used by the schemeto enhance the robustness of the watermark and to provide dual protection for the video.5.A storage centered online transaction architecture and corresponding protocolare proposed,which settle the storage problem of digital products and transactionrecords in the buyer-seller watermarking protocol.The architecture and the protocol canensure the security of the digital products and their transaction records,and make themeasy to be managed.At the same time,the boundaries between sub-protocols are clearer.6.The concept of multi-copy watermark space is given;and the model of optimalsearch for static target is improved,which is used in watermark space for setting upoptimal watermark extraction sequence.Multi-copy watermark scheme can enhance thecomprehensive robustness of the watermark,but it is uncertain which watermark copyexists after attacks,so the extraction time is quite long for an exhausting search.Theoptimal extraction sequence can make the expected extraction time minimal.更多还原