Architectures for real-time volume rendering

Hanspeter Pfister
1999 Future generations computer systems  
Over the last decade, volume rendering has become an invaluable visualization technique for a wide variety of applications. This paper reviews three special-purpose architectures for interactive volume rendering: texture mapping, VIRIM, and VolumePro. Commercial implementations of these architectures are available or underway. The discussion of each architecture will focus on the algorithm, system architecture, memory system, and volume rendering performance. Color pictures are available at
more » ... :// Introduction Visualization of scientific, engineering or biomedical data is a growing field within computer graphics. In many cases, the objects or phenomena being studied are volumetric, typically represented as a three-dimensional grid of volume elements, or voxels. Examples of volume data include 3D sampled medical data (CT, MRI), simulated datasets from computational fluid dynamics, or computed finite element models. One of the key advantages of volumetric data is that, unlike surface-based representations, it can embody interior structure of the objects. Additionally, operations such as cutting, slicing, or tearing, while challenging for surface-based models, can be performed relatively easily with a volumetric representation [Gib97]. Volume rendering generates images directly from the volume data without intermediate surface models. It allows the display of internal structures, including amorphous and semi-transparent features. Voxels are either processed in image-order or object-order to generate an image. Image-order algorithms iterate over all pixels of the output image and determine the contributions of voxels towards each pixel. Raycasting is the most commonly used image-order technique [Lev88] . Rays are cast Preprint submitted to Elsevier Preprint
doi:10.1016/s0167-739x(98)00051-x fatcat:3gmfsyjv7nggnlfvkoltst6lcy