【作者】 王启超；

【导师】 王勋；

【作者基本信息】 浙江工商大学 ， 计算机软件与理论， 2011， 硕士

【摘要】 随着数据采集技术、计算机硬件以及软件的发展,数字多媒体数据在经历—维声音、二维图像、二维视频为表现形式之后,出现了第四代数字多媒体形式---三维数字几何模型。由于其具有强烈的真实感,更符合人类的直观等优点,在影视制作、网络游戏、计算机动画、工业设计、科学计算和文物保护等领域得到了广泛应用。本文以计算机图形学、微分几何学、计算几何和数字信号处理为坚实的理论基础,对现有的网格变形技术和纹理优化技术进行了总结。针对网格模型的主要变形技术和纹理化模型的纹理空间优化所存在的问题,提出相应的处理方法。主要内容包括：1)提出了一种基于四面体胞的尽可能刚性网格变形算法。算法首先对网格模型内部进行稀疏四面体化,再通过最小化相应的变形能量函数,得到所需的模型变形结果。同时,针对大尺度编辑可能造成的模型局部塌陷,提出一种简单高效的四面体胞自适应剖分方法,进而消除不正确的变形效果。实验结果表明,该变形算法可以有效保持模型的表面细节以及模型的内部体积,并能够避免模型形状在大尺度变形时的局部退化。2)提出了一种基于视觉重要性特征图的纹理空间优化算法。算法首先通过考虑三维模型的几何信息和模型对应的纹理图片的视觉属性,计算得出对应模型的重要性特征图。然后,根据模型重要性特征图,对纹理进行松弛操作,以重新分布图像中的信号信息。同时,针对松弛过程中出现的网格严重扭曲现象,提出一种局部最优化形状匹配算法,该算法通过保持网格局部变化区域的刚性,来调整控制网格在大变形时产生的扭曲现象。实验结果表明,该变形算法可以有效节省纹理空间,并能够有效避免映射过程中产生的不连续、走样现象。更多还原

【Abstract】 Due to the rapid development of data acquisition, computer hardware and software,3D digital geometry models have emerged to be the fourth generation of digital media which had experienced several representations, such as 1D sound,2D image and 2D video. Because of the advantages of high reality and more human-satisfied intuition, this kind of representation has been widely used in those, fields, like video production, network game, computer animation, industry design, scientific computing and cultural relics protection.Based on the fundamental theories of computer graphics, differential geometry, computational geometry and digital signal processing, the intensive review and comprehensive analyses of various approaches to mesh deformation and texture optimization are presented. Moreover, we propose some solutions aiming at solving problems related to mesh model deformation and texture space optimization. Main contributions of the dissertation include:1) A novel approach to as-rigid-as-possible mesh deformation based on tetrahedral cells is introduced. First, sets of tetrahedral cells are produced through sparse voxelization. Preserving the rigidity of local transformations of mesh surface and these tetrahedra during deformation is achieved by minimizing the corresponding energy formulation that prevents unnatural artifacts both in the surface and interior of the geometry model. Meantime, in order to avoid the possible collapse emerged under large-scale deformation, we present a simple adaptive tetrahedron decomposition method. In our experiments, we show that the volume and surface details of the whole mesh can be approximately preserved and the local degeneration inside the mesh under large-scale deformation could be efficiently avoided.2) A texture space optimization algorithm based on perceptual saliency feature map is proposed. First, the corresponding saliency map of textured model is efficiently computed by combining both geometry and texture properties. Then the texture image is highly relaxed according to this saliency map. In other words, the essential signals in the texture image are accordingly redistributed. At the same time, an optimization algorithm is presented to deal with the phenomenon of severe mesh distortion. By preserving the rigidity of local mesh, this algorithm could adjust the positions of control mesh during large-scale deformation and then avoid artifacts when mapped back to the model.更多还原