Sammendrag
Abstract
Line Integral Convolution (LIC), introduced in 1993 by Cabral and Leedom [1], is a powerful texture-based
technique for vector field visualization. Based upon blurring an input texture locally along field lines
of a given vector field, it generates images that convey the directional information of the field. Due to
the computational expense of generating the 3D textures and the difficulties of effectively displaying
the resulting output textures, the application of LIC has most commonly been to depict vector fields in
2D or over a surface in 3D. We propose and study methods and strategies for more efficient visualization
of three-dimensional vector fields with LIC.
For large 3D data sets (512^3 or greater), even the Fast LIC algorithm proposed by Stalling and Hege [2]
is very computationally intensive. We present a new method for computing 3D LIC textures called Seed LIC.
This method exploits the sparsity of the input texture by calculating field lines and computing the
convolution starting from a set of distributed points (the seed points) only. The seed points can be
chosen using certain properties of the field to be visualized, e.g. the vector magnitude. Sparse input
textures have been used together with LIC to reveal the interior structures of the field [3]. The Seed
LIC algorithm can be faster than the Fast LIC algorithm by more than an order of magnitude, and can thus
be used in an interactive setting. Instead of a computation time of several hours, the computation time
for generating the LIC textures can be reduced to a few minutes or even seconds. Earlier, 3D LIC have
been thought of as a tool only to visualize vector fields in those instances where high image quality is
desired and the ability to generate images at an interactive rate is not required.
To reveal the depth relation among the eld lines, a shading, or halo, technique called limb darkening
is used. To emphasize the halo effect and to reduce aliasing effects due to the use of voxels to represent
the field lines, the textures are oversampled by a factor of 2 to 4 in each direction and then convolved
with an isotropic 3*3*3 filter. This approach smears the field lines outwards, making the strokes in the
output texture thicker and the 3D shape more clear.
Finally, we show how direct volume rendering applied to 3D LIC can be used to display additional
information about related scalar quantities when visualizing vector elds. This includes the presentation
of a new two-field visualization technique, which we have called candy cane visualization. We use this
technique to depict the directional structures of the vorticity field inside vortices obtained from a
simulation of strati ed shear turbulence by Werne and Fritts [4].
References
[1] B. CABRAL and C. LEEDOM. Imaging Vector Fields Using Line Integral Convolution. In Computer Graph-
ics Proceedings, volume 27 of Annual Conference Series, pages 263-270, July 1993.
[2] D. STALLING and H.-C. HEGE. Fast and Resolution Independent Line Integral Convolution. In Computer
Graphics Proceedings, Annual Conference Series, pages 249-256, August 1995.
[3] V. INTERRANTE and C. GROSCH. Recent Advances in Visualizing 3D Flow with LIC. ICASE Report No.
98-26, NASA Langley Research Center, July 1998. http://citeseer.nj.nec.com/interrante98recent.html.
[4] J. WERNE and D. C. FRITTS. Stratified shear turbulence: Evolution and statistic. Geophysical research
letters, 26(4):439-442, February 1999.