【作者】 郭勇；

【导师】 顾力栩；

【作者基本信息】 上海交通大学 ， 计算机应用， 2007， 硕士

【摘要】 在近十几年时间里,越来越多的科研人员将精力投入到以点为基本图形单元的图形绘制和等值面提取算法的研究上,从而形成了点图形学这一新兴领域。这种新的方法已经逐渐显示出了其在科学计算可视化领域的重要地位,同时也显示出了其在等值面提取中的存储效率和计算速度上的巨大优势。同时,随着图形卡硬件的发展和其可编程性能的进一步完善和提高,基于图形卡的通用计算已成为热门研究课题。利用成熟的图形卡编程技术,我们不仅可以大大优化和提高与计算机图形显示有关的问题,甚至可以使用它代替CPU进行大量的一般意义的计算。图形卡处理器通用计算广泛应用于图形加速显示,计算机游戏,数字图像处理,计算机仿真,科学计算,医疗图像处理系统,关键字查找,数学计算等众多领域。而今,许多相关的图形卡芯片制造厂商越来越重视图形卡作为通用计算器的设计和开发。本文也将GPU对等值面提取加速这一研究纳入其中。文章将重点论述以点为基本图形单元并由GPU进行运算的等值面提取算法的研究,主要应用于医学图像处理中的三维重建工作。以下是总结文章提出的算法的创新点。首先,传统等值面提取算法提取到的等值面基本图形单元是三角片。每个三角片由3个顶点组成,每个顶点有一个法向量。这样,在32位机器上将为每个三角片付出大小为48 Bytes的代价,而通过以点近似三角片,我们只需付出24 Bytes的代价,从而在绘制速度上和存储开销上有更多还原

【Abstract】 In recent years, more and more researchers dedicate their efforts to the research of Point Based rendering technology and Isosurface extraction algorithms. This trend makes a new field of Point Based Computer Graphics which has proved to be of great importance in“Visualization in Scientific Computing”and has great potential in the storage efficiency and computing speed of Isosurface extraction.Meanwhile, with the great improvement in graphics’programmability, general purpose computing on graphics processors (GPGPU) is becoming a hot research field. Based on the advanced graphics’programmability, we can use graphics to solve the problems not only related to the display but also related to common calculating in stead of CPU. Now, GPGPU is being used widely and is greatly employed in the fields such as visualization, PC games, digital image processing, and emulator, scientific calculating, data searching and so on. Based on the rapid development of GPGPU, now, more and more graphic companies are paying much attention to the characters design, which can support more and more GPGPU programming features. So this thesis will include the implementation of GPU’s great computing power to optimize our new algorithm.In the thesis, we will focus on the research of a Point-Based GPU accelerated Isosurface extraction algorithm which is mainly used in 3D reconstruction of medical image. The following parts illustrate the innovations:First, the traditional Isosurface extraction algorithms’result is represented by a huge number of triangles. Every triangle has three vertexes and three更多还原