Elsevier

Computers & Graphics

Volume 27, Issue 5, October 2003, Pages 725-734
Computers & Graphics

Graphics hardware
Recent advances in hardware-accelerated volume rendering

https://doi.org/10.1016/S0097-8493(03)00146-8Get rights and content

Abstract

The programmability and texture support of consumer graphics accelerators have drawn a lot of attention from visualization researchers, resulting in some very important advances in interactive volume data visualization. For many applications, scientists can now perform routine data visualization and analysis tasks on their desktop PC with a consumer graphics card that was designed mainly for playing video games. This paper presents several representative hardware-accelerated algorithms that have been introduced recently to address the problems of classification, illumination, non-photorealistic rendering, decoding, and image compositing in volume data visualization.

Introduction

Volume rendering is a powerful technique for visualizing sampled data describing physical phenomena or structures in 3-space, from molecular structures and dynamics, neuron structure, the anatomy of the human body, chemical reactions inside a furnace, air flow surrounding a vehicle, ocean temperature distribution, to the birth of our solar system. The advent of hardware support for real-time volume rendering has made this 3-D rendering technique even more attractive for a growing range of applications. Notable examples include SGI RealityEngine's support of texture mapping [1], the VolumePro volume rendering acceleration board [2], and reconfigurable volume rendering systems such as VIZARD II which uses FPGA for fast design changes [3]. More recently, driven by the video game industry, advances and innovations in consumer graphics hardware have been made at a rather fast pace. The low cost and high performance of the consumer graphics cards have led to many creative uses of several advanced features, such as high precision arithmetic and programmability, for volume graphics applications.

This paper gives an overview of the evolving commercial hardware support for volume rendering and samples of representative research results in advancing the art of hardware-accelerated volume rendering. Specifically, we describe how graphics accelerators can be used to assist volume classification, how illumination and non-photorealistic rendering (NPR) can be added to increase the clarity of the visualization, how to accelerate the rendering of time-varying data, and different hardware options for the construction of a cluster of PCs for interactive volume graphics applications. Finally, we suggest directions for further research.

Section snippets

Texture hardware features and volume rendering

Volume rendering involves resampling, classification, shading, and compositing. The latest consumer graphics cards designed mainly for playing video games can accelerate almost all the volume rendering calculations, making possible real-time rendering rates.

With texture hardware support, volume rendering is done by drawing a set of view-aligned polygon slices that sample a 3-D texture containing the volume data, as shown in Fig. 1. These slices are composited using hardware alpha-blending to

Classification

In volume rendering, each voxel must be first mapped to a color and an opacity value before the projection and compositing calculations are done. This mapping is equivalent to a classification of the volume. There has been a great deal of research devoted to the generation of transfer functions for volume classification [5].

A straightforward method for implementing a 1-D transfer function that maps each scalar value to a color and opacity is to use paletted textures, with the palette consisting

Lighting

Lighting can greatly increase the visual quality of volume rendered images by providing subtle depth cues and feature highlighting. To include lighting, however, normalized gradient direction of each voxel must be either pre-computed and stored or calculated on the fly. Rather than saving the normal vector for every voxel which would significantly increase the storage requirements, we can use paletted textures and store for each texel an index into a normal lookup table, with each direction

Non-photorealistic rendering

NPR can be used to illustrate subtle spatial relationships that might not be visible with more realistic rendering techniques. NPR for volumetric data visualization has recently become an area of active research. Treavett and Chen [10] show how pen-and-ink rendering can be applied to volume visualization. Ebert and Rheigans [11] describe a number of NPR techniques that can be applied to volume rendering. They show that non-photorealistic methods can enhance features and improve depth

Rendering isosurfaces from 3-D texture

Isosurface visualization is widely used in many engineering and medical applications. Isosurfaces are mostly extracted in a view-independent manner and represented as triangular meshes that can be efficiently rendered with polygon graphics hardware. The amount of time required for isosurface extraction is heavily dependent on which isovalue is used, where some isovalues can result in extremely dense geometry and thus high polygon counts, hampering interactive rendering. Using texture hardware,

Large volume data

The size of the volume that can be rendered interactively is limited by the amount of video memory the graphics card contains. The sheer size of a data set from a contemporary scientific application can easily overwhelm a commodity graphics card which typically has up to 256MB. For data too large to completely fit in the video memory, the rendering performance is thus limited by how fast data can be transferred from the main memory to the video memory.

Volume rendering clusters and image compositing hardware

Parallel volume rendering offers a feasible solution to the large data visualization problem by distributing both the data and rendering calculations among multiple computers connected by a network. In sort-last parallel volume rendering, each processor creates an image of its assigned subvolume, which is blended together with other images to derive the final image. Improving the efficiency of this compositing step, which requires interprocessor communication, is the key to scalable,

Conclusion

We have presented a variety of ways to exploit the high precision, programmable, texture mapping features of consumer graphics hardware for advancing the state of the art in volume data visualization. At the time of this writing, the representative consumer graphics cards on the market are nVidia GeForce FX and ATI Radeon 9800 Pro. New graphics cards which offer more video memory and programmable features are made available every year. It is thus difficult to exploit the full potential power of

Acknowledgements

This work has been sponsored in part by the US National Science Foundation under Contracts ACI 9983641 (PECASE award) and ACI 9982251 (the LSSDSV program); the US Department of Energy under Memorandum Agreements No. DE-FC02-01ER41202 (SciDAC program) and No. B523578; the National Institute of Health through the Human Brain Project; and a United States Department of Education Government Assistance in Areas of National Need (DOE-GAANN) Grant P200A980307. The confocal microscopic data set was

References (21)

  • Cabral B, Cam N, Foran J. Accelerated volume rendering and tomographic reconstruction using texture mapping hardware....
  • Pfister H, Hardenbergh J, Knittel J, Lauer H, Seiler L. The VolumePro real-time ray-casting system. In: Proceedings of...
  • Meißner M, Kanus U, Wetekam G, Hirche J, Ehlert A, Straßer W, Doggett M, Forthmann P, Proksa R. VIZARD II: a...
  • Hadwiger M, Theußl T, Hauser H, Gröller E. Hardware-accelerated high-quality filtering on PC hardware. In: Proceedings...
  • H. Pfister et al.

    The transfer function bake-off

    IEEE Computer Graphics and Applications

    (2001)
  • Engel K, Kraus M, Ertl T. High-quality pre-integrated volume rendering using hardware-accelerated pixel shading. In:...
  • J. Kniss et al.

    Multidimensional transfer functions for interactive volume rendering

    Transactions on Visualization and Computer Graphics

    (2002)
  • Tzegn F-Y, Lum E, Ma K-L. A novel interface for higher-dimensional classification of volume data. In: Proceedings of...
  • Van Gelder A, Hoffman U. Direct volume rendering with shading via three-dimension textures. In: ACM Symposium on Volume...
  • Treavett S, Chen M. Pen-and-ink rendering in volume visualisation. In: Proceedings of IEEE Visualization 2000...
There are more references available in the full text version of this article.

Cited by (10)

  • A novel software platform for medical image processing and analyzing

    2008, IEEE Transactions on Information Technology in Biomedicine
  • From volume splatting to raycasting - A smooth transition

    2008, GRAPP 2008 - Proceedings of the 3rd International Conference on Computer Graphics Theory and Applications
  • Octree based representation and volume rendering of three-dimensional medical data sets

    2008, BioMedical Engineering and Informatics: New Development and the Future - Proceedings of the 1st International Conference on BioMedical Engineering and Informatics, BMEI 2008
View all citing articles on Scopus
View full text