Sign up & Download
Sign in

ShellSplatting: Interactive Rendering of Anisotropic Volumes

by Charl P Botha, Frits H Post
Data Visualization 2003 Proceedings of Joint Eurographics IEEE TCVG Symposium on Visualization (2003)

Abstract

This work presents an extension of shell rendering that is more flexible and yields higher quality volume renderings. Shell rendering consists of efficient data-structures and methods to manipulate and render structures with non-precise boundaries in volume data. We have updated these algorithms by creating an implementation that makes effective use of ubiquitously available commercial graphics hardware. More significantly, we have extended the algorithm to make use of elliptical Gaussian splats instead of straight-forward voxel projection. This dramatically increases the quality of the renderings, especially with anisotropically sampled volumes. The use of the graphics hardware alleviates the performance penalty of using splats.

Cite this document (BETA)

Available from www.eg.org
Page 1
hidden

ShellSplatting: Interactive Rendering of Anisotropic Volumes

Joint EUROGRAPHICS - IEEE TCVG Symposium on Visualization (2003)
G.-P. Bonneau, S. Hahmann, C. D. Hansen (Editors)
ShellSplatting: Interactive Rendering of Anisotropic Volumes
Charl P. Botha and Frits H. Post
Data Visualisation Group
Delft University of Technology, The Netherlands
{c.p.botha,f.h.post}@its.tudelft.nl
http://visualisation.tudelft.nl/
Abstract
This work presents an extension of shell rendering that is more flexible and yields higher quality volume ren-
derings. Shell rendering consists of efficient data-structures and methods to manipulate and render structures
with non-precise boundaries in volume data. We have updated these algorithms by creating an implementation
that makes effective use of ubiquitously available commercial graphics hardware. More significantly, we have ex-
tended the algorithm to make use of elliptical Gaussian splats instead of straight-forward voxel projection. This
dramatically increases the quality of the renderings, especially with anisotropically sampled volumes. The use of
the graphics hardware alleviates the performance penalty of using splats.
1. Introduction
Volume visualisation can be performed in three ways: ren-
dering two-dimensional slices of the volume, rendering sur-
faces that have been extracted from the volume and direct
volume rendering1.
An extracted surface is usually an isosurface and is ap-
proximated as a triangulated mesh for accelerated render-
ing with modern polygon graphics hardware. This method
assumes that extractable isosurfaces are present in the data
and that these isosurfaces correctly model the structures in
the volume2.
Direct volume rendering3, 4 (DVR) allows us to visu-
alise structures in the data without having to make deci-
sions about the precise location of object boundaries by ex-
tracting polygonal isosurfaces. Instead, we can define multi-
dimensional transfer functions that assign optical properties
to each differential volume element and directly visualise
structures on the grounds of this transformation.
Shell rendering5, 6, which can be seen as a hybrid of sur-
face and direct volume rendering, was proposed as a fast vol-
ume visualisation method in the early nineties. It had sig-
nificant advantages at that time: it was an extremely fast
software algorithm that also required far less memory than
competing volume visualisation algorithms. As recently as
2000, software shell rendering was still found to be faster
than hardware assisted isosurface rendering7. Originally it
supported only parallel projection but work has been done
to extend it to perspective projection8. In this paper, we will
consider only parallel projection.
However, this object-order method makes use of simple
voxel projection to add a voxel’s energy to the rendered im-
age. In addition, texture-mapping, shading and composit-
ing hardware has become common-place. These two facts
have led us to extend shell rendering by translating it to
a hardware-accelerated setting and to allow voxels to con-
tribute energy to the rendered image via a splat-like ellip-
tical Gaussian. This makes possible interactive high-quality
renderings of anisotropically sampled volumes.
In this paper we present this extension that we have
dubbed ShellSplatting. The algorithm generates higher-
quality renderings than shell-rendering and does this more
rapidly than standard splatting. It retains all advantages
of the shell rendering data-structures. We also present a
straight-forward way to calculate the splat projections that
accommodates anisotropically sampled volumes at no extra
speed or memory cost.
Section 2 supplies information about volume visualisa-
tion, focusing on splatting and shell rendering and the de-
velopment of this work. In section 3 our algorithm is doc-
umented. Section 4 contains visual results as well as com-
parative timings. In section 5 we detail our conclusions and
mention possible avenues for future work.
c© The Eurographics Association 2003.
105

Sign up today - FREE

Mendeley saves you time finding and organizing research. Learn more

  • All your research in one place
  • Add and import papers easily
  • Access it anywhere, anytime

Start using Mendeley in seconds!

Already have an account? Sign in

Readership Statistics

3 Readers on Mendeley
by Discipline
 
 
by Academic Status
 
33% Student (Bachelor)
 
33% Ph.D. Student
 
33% Assistant Professor
by Country
 
33% South Korea
 
33% Netherlands
 
33% Brazil