【作者】 孟鑫；

【导师】 张桦；

【作者基本信息】 天津理工大学 ， 计算机应用技术， 2006， 硕士

【摘要】 本文对体绘制技术进行了深入研究,介绍了它的产生背景和应用领域。与传统的面绘制技术相比,体绘制技术具有明显的优势,因此成为计算机图形学的一个发展热点。由于体绘制技术本身需要对大规模的数据场进行重构处理,算法执行速度较慢,这成为其发展的瓶颈。目前该领域的主要研究方向集中在改善绘制图像的质量和提高渲染速度两个方面上,论文首先介绍了当前国内外学者对体绘制技术的研究成果和该领域的整体水平,随后提出一种采取多重构核的方法改进Splatting算法的图像绘制效果,接着运用并行计算技术对体绘制算法进行加速,最后得出结论。Splatting算法作为物序体绘制算法的代表,具备图像质量高、执行速度相对较快以及参加计算的体素可以灵活选择等优点。本文对该算法进行了深入研究,对Splatting的核心内容──重构核进行了详细分析,得出选择重构核的标准;提出一种采取多重构核的Splatting体绘制算法,对数据场内的不同体素进行分类后,应用以不同的重构核,利用不同重构核的特征计算其对最终图像的影响。由于重构核可以在算法执行前采取预计算方式得到,因此该方法在没有大幅度增加绘制时间的前提下,提高了图像的质量,使结果图像中包含更多的细节信息。本文针对基于工作站机群的并行体绘制算法进行了研究。利用现有PC工作站通过高速局域网组成的并行环境以低廉的成本实现大型机的计算效果,而在不执行并行任务时,集群环境内的计算机仍然可做它用。论文采取消息传递机制作为体绘制算法并行化的思路,设计并实现了一个基于Windows平台的并行Splatting程序。该并行程序采取主从模式,使用动态任务分配策略,保证负载均衡。实验结果证明,该并行体绘制程序实现了对串行算法加速的结果,适合于大规模数据场的绘制。最后,对本文的研究内容进行总结,并指出未来继续进行研究的方向。更多还原

【Abstract】 The diploma thesis firstly describes the background and successful fields of volume rendering, then makes a thoroughly study on its popular algorithms. Compare to traditional surface representation direct volume rendering technique has got its own advantages, but because of the large size of volume data the rendering speed becomes the bottleneck of volume rendering. The main research fields of volume rendering are reducing time the whole process consumed and improving the image quality. After comprehending the domestic and international achievements of the volume rendering, we present a new method using multi reconstruction kernels to improve image quality of splatting algorithm. A high-quality parallel volume visualization system based on cluster of workstations has been developed which gives a speedup of the volume rendering process.Splatting algorithm, relying on its high picture quality, relatively high-speed and ability to freely choosing voxels become the best one in Object-Order volume rendering. We carry out an in-depth study on splatting. Due to its mathematics principle, reconstruction kernels are the key of the algorithm, we introduce the criterion to choose proper reconstruction kernels. We present a new method for splatting process, different reconstruction kernels are applied according to the property of voxels. As a result, the outcome picture will have more details than before. Since the reconstruction kernels can be computed before the rendering process, the new method can achieve the better image quality and don’t obviously increase the rendering time.The cluster of workstation model aims at using the existent hardware via high-speed network structure to achieve the capability of mainframe. The thesis makes use of message passing mechanism to develop a parallel volume rendering. The design bases one the Windows system and takes the master-slave task distribute model. Using a dynamic task assign method ensure the whole parallel computing platform get load-balance. The experimentation proved that the parallel mechanism we designed have speedup the volume rendering process. It especially adapts to the large-scaled volume rendering task.At last, we summarizes the thesis, points out the future explore directions.更多还原