【作者】 李震；

【导师】 陆哲明；

【作者基本信息】 哈尔滨工业大学 ， 仪器科学与技术， 2006， 硕士

【摘要】 伴随着网络技术和多媒体技术的飞速发展,如何保护多媒体信息的安全成为国际上研究的热门课题。三维网格模型作为一种新型多媒体产品得到日益普及和广泛应用,它的版权保护与内容认证成为人们日益关注的热点。同时,随着三维网格数据量的迅猛增长,压缩域三维网格数据的处理十分必要。本论文是关于三维网格模型水印技术与三维网格模型的压缩技术的一些研究,研究成果包括:系统地介绍了三维网格模型数字水印技术的发展现状,介绍了一些典型的三维网格模型数字水印算法,并重点介绍了可逆水印技术的研究现状;简要介绍了三维网格模型压缩技术基本概念、研究背景和重要意义以及发展现状并着重介绍了压缩域水印技术的发展。旨在了解现有水印算法、压缩算法发展进程,为本文水印、压缩以及两者结合的算法的提出提供思路。提出了一种三维网格模型鲁棒水印算法:以三维网格模型为研究对象,利用模型顶点的几何坐标作为水印嵌入基元,引入了电路理论中的并联阻抗的性质,使水印的嵌入强度自适应于顶点的局部几何特征。同时,与已往自适应水印嵌入算法不同的是:水印嵌入方向也自适应于局部几何特征。这使得本算法能够在嵌入的水印信息不可见的前提下嵌入更大能量的水印,从而保证水印隐藏对象能够承受更大强度的攻击。实验表明该算法鲁棒性较好,对噪声、旋转、平移、缩放、剪切、降采样及其联合攻击的抵抗性较强,并且算法复杂度较低。提出了两种三维模型矢量量化压缩算法:一种是基于扩展码书的,它能够充分利用三维网格中相邻顶点的码字索引的相关性,将邻域顶点的码字索引存储到FIFO中,使码字搜索范围减小,从而获得更低的比特率,同时节省了码字搜索时间;另一种为基于k环预测的,它能利用顶点的k环邻域内的所有顶点的预测误差估计得到待量化顶点的预测误差,从而进行预测误差补偿,减小了残差矢量的动态范围,进而使网格模型在相同的比特率的情况下进行矢量量化压缩造成的失真更小。仿真结果表明,这两种算法均获得比经典算法更高的比特率-失真性能。基于目前大多数可逆水印算法针对原始域媒体的不足,结合矢量量化的思想,提出了一种压缩域三维网格可逆数字水印算法:该算法通过改变矢量量化压缩过程的矢量预测方式来嵌入水印信息,使水印对于信道噪声引起的更多还原

【Abstract】 The multimedia information security problems including copyright piracy emerged as side effects over network. 3D mesh models are being used widely as a new type of media, and the problems of copyright protection and content authentication have become the hotspots of 3D mesh processing. Meanwhile, with the rapid growth of the data volume of 3D mesh models, it becomes necessary to process 3D data in the compressed domain. In this dissertation, we study the technologies of 3D model watermarking and compression. The main contributions of this dissertation are:Firstly, state of art watermarking technologies for 3D mesh models, especially those of reversible watermarking, and some typical techniques are systematically introduced here; we also briefed the basic concepts, research background, significance and state of art technologies of 3D mesh models compression, including some developments of watermarking techniques in the compressed domain. The above summarizations are aimed to provide new thoughts in the areas of watermarking, compression and their combination.Secondly, a robust watermarking scheme for 3D mesh models is proposed: we select vertices’coordinates as primitives for embedding watermark. By introducing the parallelogram impedance property here, we make the watermark embedding intensity adaptive to the local geometry characteristic of the vertex. At the same time, the direction along which the watermark bit is embedded is also adaptive to the local geometry, which is rather a novel idea. The proposed adaptive watermarking scheme enables more energy of watermark to be embedded with imperceptibility, so the stego mesh can be resilient to higher intensity of attacks than with traditional schemes. The experiment results show that the watermark embedded is rather robust to various attacks such as adding noise, rotation, translation, scaling, insecting, down-sampling and some of their combinations. Also the computation complexity is rather low.Thirdly, we proposed two algorithms for 3D mesh models compression using vector quantization: one scheme is based on an extended codebook, with which the correlations between indices of adjacent vertices are adopted and更多还原