Graphics Reference
In-Depth Information
method was based on his earlier work in representing and rendering objects as a
collection of scattered points. It is said that this technique was the initial idea for
the volume rendering technique.
Active research into volume rendering was being done at Pixar in the mid-
1980s. A technical report written by Alvy Ray Smith at Pixar was said to be
the first volume-rendering paper in CG; however, it was not made public at the
time [Smith 87]. The original goal was to develop a commercial medical imaging
system; the volume visualization research team at Pixar included the now famous
radiologist Elliot Fishman. The system was completed and released, but its high
cost and the fact that it was perhaps too technologically advanced for its time
prevented it from having much success as a commercial product. The work was
eventually summarized in a SIGGRAPH paper entitled “Volume Rendering” au-
thored by Robert A. Drebin along with Loren Carpenter and Pat Hanrahan [Drebin
et al. 88]. This paper introduces a highly accurate method for rendering images
of volumes in CT data that is particularly suited for imaging the human body.
The ideas represented a significant leap forward in volume rendering and greatly
influenced subsequent research, in both volume rendering in general and the vi-
sualization of the human body based on scientific data.
A great advantage of the volume rendering technique presented in this paper is
its speed. The algorithm is also useful for rendering both isosurfaces and volumes,
and is well suited to hardware rendering. For these reasons it can be said that the
algorithm is the most suitable technique for real-time medical imaging. Its use
in imaging the human body is described in the next subsection. The approach is
also useful for rendering environments that contain participating media. Realistic
simulation of smoke, fog, and clouds in fluid dynamics typically use voxel-based
finite-element methods. Discrete-time simulation stores the current fluid speed,
pressure, and density in voxels at each time step. When these methods are applied
in CG, the same voxels can be used for volume rendering.
3.2.3 Volume Rendering of the Human Body
Figure 3.5 illustrates the algorithm described in the SIGGRAPH “Volume Ren-
dering” paper [Drebin et al. 88]. Each slice of data captured by a CT scan is
processed in parallel, and then the results are synthesized into the final 3D render-
ing. The approach was originally developed to visually reconstruct the internal
structure of the human body from the complex CT data, but it can be applied to
other volume data sets.
One of the most significant advances was the method of handling region
boundaries without explicitly reconstructing the boundary surfaces. Before that,
most visualization techniques required a geometric representation of the surface
