【作者】 金磊；

【导师】 董军宇；

【作者基本信息】 中国海洋大学 ， 计算机系统结构， 2010， 硕士

【摘要】 过程纹理生成一直是计算机虚拟现实领域中一个至关重要的问题,它主要用于模拟自然界中常见的大理石、云朵、树木表皮等纹理。大多数的过程纹理都是基于某类噪声函数的,本文采用的被广泛应用的Perlin噪声函数是其中最强大的一种。在过去,由于过程纹理计算量很大,在实时绘制中很少使用。但是GPU的出现,促进了过程纹理在实时渲染中的广泛应用。目前噪声函数作为生成过程纹理的重要工具主要针对二维的过程纹理生成,三维过程纹理的生成仍在有待探索。纹理分类是涉及数字图像处理、机器学习、模式识别等多门学科的内容的热点问题,它的目的是识别出具有不同纹理特征的物体的各种表现形式。纹理分类在计算机视觉、图像处理及计算机图形学领域和工程技术方面有着非常广泛的应用背景。分类的方法主要包括使用聚类的方法、基于机器学习的方法等。目前三维表面纹理分类技术的研究是图像处理领域的前沿方向,有着广阔的研究前景。本文首先介绍了虚拟现实中的过程纹理生成,以及模式识别领域中纹理分类的几种方法和相关背景研究,进而展开对三维表面纹理的生成和分类研究。在前期Perlin噪声函数生成的Cellular Texture的基础上,运用GPU编程技术,提出了三维过程纹理的构造方法。并且基于朗伯模型将生成结果,应用到三维纹理重光照中。针对Cellular Texture构造函数的结构特点,本文使用函数参数的组合系数作为表达纹理的特征向量,实现了基于SVM和聚类分析的两种分类器的三维表面纹理分类算法,并进行了大量的实验。实验结果证明本文提出的三维表面纹理生成和分类算法是行之有效的,并为今后进一步的研究打下了一定的基础。更多还原

【Abstract】 Procedural texture generation has been a critical issue in the field of computer virtual reality, which is mainly used to simulate the common nature marble, clouds, trees and other skin texture. Most of the procedural texture functions are based on certain types of noise. Perlin noise used in this paper is the most powerful one, which is widely used. In the past, the process of texture is rarely used in real-time rendering for its large number calculation. However, the emergence of GPU promotes wide applications of procedural texture in real-time rendering. Recently, noise function is mainly directed as an important tool for generating procedural texture towards two-dimensional aspect, three-dimensional procedural texture generation is still waiting to be explored.Texture classification is a hot issue related to digital image processing, machine learning, pattern recognition and many other subjects. The purpose is to recognize various manifestations of an object, which have different kinds of texture features. Texture classification has broad applications in computer vision, image processing, computer graphics and engineering technology, the methods of which includes clustering, machine learning, and so on. Still, three-dimensional surface texture classification study is the forefront of image processing in the field direction, and has broad researching prospects.This article firstly introduces several research methods and relevant background of texture classification in the field of pattern recognition feature extraction, and that of texture generation in the field of virtual reality, then expand to the research on three-dimensional surface texture. This paper use GPU programming to propose a construction method of three-dimensional procedural texture, such as Cellular Texture, which is the following formation of Perlin noise function. And the results based on Lambertian model will be applied to three-dimensional texture re-lighting. For the structural characteristics of Cellular Texture constructor, we treat the factors’ combination of the function arguments as an the feature vector of texture expression, to achieve surface texture classification algorithm based on classifier of SVM and on clustering analysis, finally make a lot of experiment. Experimental results show that the algorithm of the proposed three-dimensional surface texture classification and generation is effective, and set a foundation for the future research.更多还原