【作者】 黄向军；

【导师】 史烈； 陈小平；

【作者基本信息】 浙江大学 ， 计算机系统结构， 2003， 硕士

【摘要】 数字水印技术是多媒体通信和多媒体信号处理领域近年来的新的研究方向，作为多媒体作品知识产权保护的有效手段，得到非常广泛的研究和应用。三维运动作为一种新的多媒体数据内容，已成为计算机动画研究的一个热点，并在影视、游戏、控制、仿真等领域有非常广阔的应用前景。本文探讨了多媒体数据数字水印的各种典型技术和三维运动相关的理论和方法。通过对现有数字水印技术的归纳、比较和分析，结合三维运动的特点提出了两种鲁棒的盲水印算法：三维运动盲水印检测算法和三维运动盲水印提取算法。本文第一章对多媒体数据的数字水印技术进行了综述，简单介绍了数字水印的一些基本概念，阐述了数字水印的定义、产生背景、性能指标以及分类，并给 出了一个通用数字水印系统的结构。重点对现有的一些典型的水印算法做了分析，分别针对不同算法进行评价。对数字水印的攻击方法做了分类和讨论，介绍了水印技术的应用。在第二章中概述了三维运动及运动信号处理的理论和方法。论述了运动角色建模、运动数据表达的思想，然后分析了三维运动数据生成的典型方法，包括运动捕获、运动图象序列跟踪、运动学方法以及动力学方法，并对不同方法进行比较评述。阐述了运动信号处理的基本概念、理论和方法，最后介绍了我们基于OpenGL的运动数据动画显示程序。第三章基于统计检测理论与DCT变换系数分布的研究，提出一个鲁棒的三维运动盲水印检测算法，它是一个DCT变换域上的符号相关性检测算法。实验结果显示，算法对于运动自由度重排、斯白噪声、重采样以及二次水印等攻击有很好的鲁捧性。在第四章中实现了一个三维运动盲水印提取算法。本章的算法应用了量化与扩展频谱的思想，首先对运动信号进行频域变换，在DCT域上选择一定数量的中频交流系数量化嵌入水印信息。实验结果显示，我们的算法在高斯白噪声、重采样、运动光滑等攻击下有很好的鲁棒性。最后，在第五章中对全文进行归纳和总结，并且提出了进一步研究的方向。更多还原

【Abstract】 Digital watermarking tecndque is a new research area of multimediacomrnunication and multimedia signal processing in recentIy years. It’s widelyinvestigated and used as an effective way for boowledge property right protection. 3Dmotions, as a new multimedia cofltent type, have become a hotspot of computeranimation research, and also have broad aPplication prospect in movie. game’control. simu1ation field etc. This paPer discusses typical multimedia watermarkingalgorithms. theory and technique related with 3D motion. By comparing andanalyzing the existent watermarking algorithms, it proposes two robust watermarkingalgorithms f a 3D motion blind watermark detection algorithIn and a 3D motion bIindwatermark extTaction algorithIn.The chaPter 1 surveys the popular tec[mque of muhimedia watermarkng,introduces the basis of digital watermark, explains its definition. background.perfOrmance indicator. classification and aPplication field, illustrates a general digitalwatermarking system. It emphasized on the analysis and evaluation of typicalwatermarking algorithIns. It classifies and discusses the attack of watermarking,introduces the aPplication of digital watermarking tecboque in industry.In the chaPter 2, it summarizes the theory and technique of 3D motion and motionsignaI processing. It addresses the theory of character modeling’ motionrepresentation. Then analyzes and cornmeflt on the typical ways of 3D motioncreation, including motion caPtllfe’ motion pictUre sequence tracking. kinematics anddynamics. lt introduces the basis of motion signal processing, and finally is ourmotion data viewer program based on OpenGL.Based on the theoTy of statistical dctection and research on distribution of DCTcoefficiefits, the chaPter 3 proposes a robust 3D motion blind watermark detectionalgorithIn, which is a sign correlation detection scheme. Experimental results showthat this method is resilience to many attacks such as Gaussian White Noise.resaxnpling. smoothing and second watermark.In chaPter 4, it implemented a robust 3D motion blind watermark extractionalgorithm. The algorithIn adopts twO well-known ideas called spread spectrum andquantization, first applies a frequency transfOrmation on the motion signal, and thenembeds watermark bits by quantizing proper intermediate frequency coefficients. Ourexperiments show that this method is also robust to Gaussian White Noise.resampling and smoothing etc..At last, some discussions and conclusions are given.更多还原