【作者】 蒋洪波；

【导师】 卢炎生；

【作者基本信息】 华中科技大学 ， 计算机软件与理论， 2004， 博士

【摘要】 空间数据集可视化是当前计算机图形学领域的热门研究课题,它通过分析、抽象和处理表示各种物体和场景的海量数据,实现了在计算机中真实模拟现实世界的目的,因此在医学、地球科学、天文物理、化学、机械工程等许多方面有广阔的应用前景。 它是将某个物体的几何模型转变成我们视觉上清晰直观图像供进一步研究、分析,图像的绘制质量将直接影响工程技术人员对原始数据的理解,因此对可视化图形绘制的研究具有突出的现实意义。空间数据可视化绘制的主要研究内容包括光照模型、纹理技术和图形生成等。光照模型是最常用的绘制技术之一。经典的基于物理的反射模型在象素级都是一个表面几何的函数,且表示成光照明的模式,但是它不能反映随着视点的变化而产生的反射特征的变化。一种基于几何的分辨率反射模型适用于表示不同级别的反射影象,但是这种量化由分辨率来确定的并被记录在多分辨率mipmaps中,为此一种多分辨率反射技术可以用来有效记录这些基于分布的参数。该技术使用基于几何的反射函数来表示,并且通过显示硬件的支持来实现实时绘制比较好的反射细节。纹理映射和调度是可视化绘制研究的一个重要内容。以前的自动纹理映射算法存在效率低或只能基于模型特征点的约束等缺点。因此针对平面网格,可以将特征点的约束推广到基于特征曲线的约束就产生了一种新的纹理映射算法。该算法首先对每一个由待求顶点和特征曲线面上由两个构成的三角形使用共形映射计算出一个可能的纹理映射,然后使用高斯正态分布概率密度来对这些可能的纹理映射进行加权平均得到最后的结果。另外针对地形可视化中纹理映射和调度进行了改进:首先为了使得地形块边界处的纹理实现无缝连接,在数据的初始化阶段提出了一种基于重复采样的实现分块纹理无缝拼接的纹理映射算法;然后定义了纹理误差,给出了详细的数学表达式,并证明了它具有良好的性质;在此基础上综合考虑了地形纹理的屏幕投影大小和位置,给出了一个确定纹理分辨率的算法;最后提出了一种预估计的算法来控制纹理释放的调度,该算法预先调度出可能不可见的纹理块以减少内外存的数据交换。实验比较结果表明该调度算法不仅速度快而且画面质量好。模型简化是图形生成的研究重点。模型简化是图形生成的研究重点。一种新的<WP=4>度量简化误差的算法对以前的误差度量算法进行了改进,使用c-误差来度量那些在边收缩过程中顶点发生了移动的三角形面的误差,使用d-误差来度量那些在边收缩过程中退化的面产生的误差。这种将两种误差同时考虑得新的误差度量算法可以得到更高得效率和图像质量。另外,在这种误差度量的基础上提出了一种基于自适应空间刨分的模型简化算法。该算法的思路是对模型中的所有的顶点进行量化赋予一个二次误差阵,并将它们视为一个簇,沿坐标轴方向将它们刨分成八个子簇并不断迭代刨分生成新的子簇直至达到指定的精度,将最终的离散点集用适当的算法重新进行三角化得到简化模型。根据上一个算法中的自适应空间刨分思想又可以得到一个与视点相关又无须对原来的复杂模型进行预处理的简化算法,即每次绘制都根据视点位置,视方向和模型每部分产生的误差,对影响视觉效果大的部分进行自适应空间刨分从而逐步产生可显示的模型。在许多实际应用中需要保持模型的拓扑连通性,但以前的算法不可避免的会损失模型的连通信息。基于拓扑连通保持的外存模型简化算法是一种适用于任意大型和复杂物体的模型简化算法。它不仅能实现产生很简单的近似模型,而且还快速地保持了拓扑连通性。首先,算法根据自适应的栅格分辨率使用外存一致聚类的得到一些簇和误差矩阵,然后使用八叉树空间剖分和一种新的拓扑连通检查算法来保持其拓扑连通性,接着根据得到的蔟集使用顶点聚类的算法将物体简化到中等规模的简化模型,最后用面收缩的内存简化算法法得到最终简化模型。通过误差传递结合连续的几个步骤,新算法与以前的外存简化模型算法相比,可以保留更多的物体原有的拓扑连通信息。该算法被证明是正确的,同时实验结果也表明简化后模型的质量也优于其它已有的外存简化算法。更多还原

【Abstract】 Visualization for spatial dataset has received increasing attention and has reached a great achievement in the field of medicine, geoscience, astronomy, chemistry, mechanical engineering et al. It is release a large number of engineer and technologist from large scale and fussy data. Rendering technique for visualization is a fundamental tool by which the spatial dataset are outputted and transformed into meaningful visual images. The rendering quality is promising for recognition of original data to engineer and technologist. Besides, the application of graphics rendering for visualization is very wide and spreads most field of industry and society. It is foreseeable that this technique will be more and more important in the future. Visualization for rendering research covers illumination model, texture technique and graphics generation.Illumination model is one of the commonest rendering techniques. Physically-based reflectance models typically represent light scattering as a function of surface geometry at the pixel level. With changes in viewing resolution, the geometry imaged within a pixel can undergo significant variations that can result in changing reflectance characteristics. To address these transformations, we present a multiresolution reflectance framework based on geometric descriptors appropriate for representing reflectance effects over different scales. Since these quantities must be efficiently determined with respect to resolution, they are recorded at multiple resolution levels in mipmaps, for which we present a technique for efficient filtering of these distribution-based parameters. This framework can be used with common geometry-based reflectance functions, and can be implemented in graphics hardware for real-time processing. With this multiresolution reflectance technique, our system can rapidly render the fine reflectance detail that is customarily disregarded in multiresolution rendering methods.Texture mapping and caching is playing a key role in many study for visualization. Most previous automatic texture mapping algorithms have suffered from been too slowly or could only use a limit set of feature points of the model. We introduce in this paper a <WP=6>new constrained-based method for automatic texture mapping that deal with these problems for planar meshes. Our critical improvements include generalizing feature point to feature curve to fit some application and also simplified computation through efficient weighted of texture mapping. We generalize by parameterizing each vertex at two passes. Firstly we compute all the possible mapping along the feature curve consecutively where every triangle that consists of the computed vertex and two points in the curve can be deduced a conformal mapping result with minimum deformations. Then we combine all the possible mapping using normalization distribution to get the final one. Our results show that this texture mapping algorithm makes a great mapping for the planar meshes. Besides, this paper proposes a system that improves the texturing technique for large scale terrain visualization. A texture mapping method based on overlap sampling is presented for tiling textures without seams during initialization phase. And then this paper defines a texture error, presents its detail math expression and proves that it has very good character. An algorithm based on the texture error proposed to confirm texture resolution considers not only screen projection but also position. Finally advanced estimate method can control schedule for texture release that unload possible invisible texture patches to induce data commute of main and second memory. The experiments show that result has high quality with high speed.Model simplification is an important tool for graphics generation. At first, given the framework of incremental mesh simplification based on edge collapse, the paper proposes a mesh simplification algorithm using an improved approach for measuring simplification error. The algorithm uses edge collapse to simplify the triangle mesh and maintains更多还原