ERRATA
Correction: In our paper we write about [Gum03] that its "bounding box
computation is only correct for orthogonal projections, and thus might
clip or cut off ellipsoids for perspective ones." This statement is
incorrect. The method by [Gum03] DOES NOT compute an axis-parallel
bounding box and it DOES NOT cut off ellipsoids. It rather computes a
correct, perspectively distorted, general bounding quadrilateral that
contains the silhouette ellipse in world space.
The resulting quadrilateral is a tight enclosure, but in contrast to our
method it requires four times as many vertices per splatted quadric. Our
proposed rendering method uses only a single GL_POINT primitive for each
splatted quadric, requiring an axis-aligned bounding box to determine
the respective point size. Deriving such a bounding box from the general
quadrilateral bounding shape of [Gum03] would be non-trivial, as
perspective distortion has to be accounted for. The bounding box of the
quadrilateral itself, however, would be an overly conservative
estimation. Consequently, we had to develop a novel bounding box
computation.
Additional reference: Please note that our paper also relates to the
following publication:
G. Reina and T. Ertl. "Hardware-Accelerated Glyphs for
Mono- and Dipoles in Molecular Dynamics Visualization".
In Proceedings of EUROGRAPHICS - IEEE VGTC Symposium on
Visualization 2005, pp. 177--182.
The respective paper presents a molecule visualization system similar to
ours. It also employs point sprites, vertex, and fragment programs to
raycast perspectively correct spheres and cylinders. The presented
system supports depth cues and adaptive rendering. Point sizes are
computed using an unspecified heuristic.