Scalable Architecture for Real Time Volume Rendering

本研究涉及一种新的可扩展体可视化体系结构及其相关算法的开发。该架构被称为Cube-4，利用并行性和流水线来实现从体数据中实时渲染高分辨率图像。该架构基于体素体体缓冲的光线投射算法，体素体体缓冲存储为倾斜的分布式内存，以支持对体素结构的无冲突访问。它执行采样点的插值、着色和合成来生成像素值。计算是使用处理器之间有限的通信完成的，因此该架构可以在广泛的性能和图像分辨率范围内进行扩展。该项目包括算法开发、架构研究和低分辨率原型的构建。

This research deals with the development of a new scalable volume visualization architecture and its associated algorithms. The architecture, called Cube-4, exploits parallelism and pipelining to achieve real-time rendering of high-resolution images from volume data. The architecture is based on an algorithm for ray casting of a volume buffer of voxels which is stored as a skewed distributed memory to support conflict-free access to voxel structures. It performs interpolation of sampled points, shading, and compositing to generate the pixel values. Computations are done using limited communication between processors, so that the architecture is scalable over a wide range of performances and image resolutions. This project encompasses algorithm development, architecture research, and construction of a reduced-resolution prototype.