【作者】 胡健；

【导师】 李思昆；

【作者基本信息】 国防科学技术大学 ， 计算机科学与技术， 2011， 硕士

【摘要】 非结构化网格体绘制是科学计算可视化研究的重点内容之一,但非结构网格体绘制计算量大,过程复杂,因此产生图形图像速度较慢,不能满足日益增加的实时可视化要求,因此提高非结构化网格体绘制算法的效率是实现高效体绘制的一个重要研究点。本文针对传统算法图像空间片元排序效率不高的问题,研究了k-buffer处理过程及优化方法。Paraview是目前应用广泛的开源并行可视化平台,研究将数据可视化算法集成到Paraview平台上高效并行运行的方法,有利于实现大规模数据的并行可视化。本文研究了Paraview的扩展技术,总结了数据可视化算法在Paraview上实现并行可视化的扩展方法和编程规范。本文完成的主要工作及取得的研究成果总结如下:1.针对主流的HAVS算法在图像空间排序效率不高的问题,提出了针对k-buffer处理过程的优化方案,提高了绘制效率;另外,k值大小对绘制精度和效率有重要影响。已有HAVS算法将buffer存储的片元数k限定为6,制约了算法精度与性能。本文扩展了k值的范围,提高了绘制精度。2.针对Paraview并行可视化平台扩展的应用需求,提出了基于插件的并行可视化程序设计规范和方法,并通过并行体绘制算法并行化进行了验证,实现了filter插件和体绘制算法类Mapper的插件,该方法可有效指导用户根据插件扩展方法进行数据可视化算法的并行实现。3.为了进一步减少Paraview并行可视化平台的扩展工作量,研究提出了基于Paraview的并行可视化程序直接扩展方法,并通过并行体绘制算法并行化进行了验证,将非结构化网格体绘制算法HAVS通过直接扩展方法加入Paraview中,该方法为Paraview的扩展提供了一种用户可选的新途径。更多还原

【Abstract】 Improving the efficiency of unstructured grid volume rendering is an important research point of high efficient volume rendering. This paper researches on k-buffer process and optimization against inefficient fragment visibility ordering in image space.Paraview is a parallel open-source visualization application. It is benefit for parallel visualization of large scale dataset to research on integration technology of visualization algorithm to Paraview. This paper analyses the parallel frameworks of Paraview, researches on the extending technology and summarizes the extending technology and programming criterion of Paraview.The contributions and relevant work in the paper are as follows:1. At present, HAVS is one of the most important unstructured grid volume rendering algorithm frameworks based on GPU. This paper proposes an optimization algorithm against the low efficiency of sorting in k-buffer. We also test different k values and extend the range of k, which will increase the precision of the final image. The experimental results show that our optimization algorithm obviously improves the efficiency of HAVS.2. For the need of extending Paraview, this paper proposes the methods and criterion of parallel visualization program designing which based on plugin. And we verify the criterion by paralleling volume rendering algorithm and achive a filter and a mapper plugin for volume rendering. These criterions will guid users to parallel data visualization by plugin.3. For decreasing the work of extending Paraview, this paper proposes the criteria of directing extending Paraviewfor parallel viualization. We verify the criterion by paralleling unstructured grid volume rendering algorithm, which gives users a new way to extend Paraview.更多还原