Graphics Reference
In-Depth Information
Figure 10.15.
Object precision rays for volume
rendering.
Figure 10.16.
Shear-warp volume rendering.
density values at these voxels to come up with a single value at the pixel is called
com-
positing
. From a theoretical point of view, one should integrate the intensities along
rays, but different compositing functions can be chosen. The simplest is to project the
maximum density along the ray to the pixel. This is often useful when trying to isolate
well-defined structures in the view volume. We can do the compositing in a front-to-
back or back-to-front manner. One has the option of computing with the transparency
t associated to a voxel or its
opacity
o = 1 - t.
[LacL94] describes an efficient variant of ray casting called
shear-warp volume
rendering
. Rather than sending out rays that are skew to the voxel volume (Figure
10.16(a)), they apply a shear transformation to the voxels to achieve an equivalent sit-
uation where rays are cast that are perpendicular to the view plane (Figure 10.16(b)).
In this way one can define an efficient ray traversal. Two other methods used in the
algorithm to speed things up is run-length encoding of the voxel scan lines (a defini-
tion of run-length encoding can be found in [Salo99], for example) and early ray ter-
mination, where one stops the compositing process when a pixel has reached full
opacity.
Data classification in direct volume rendering amounts to defining the opacity or
transparency of voxels. This is accomplished by defining what is called a
transfer func-
tion
that maps densities, or densities together with gradient information, to opacities.
Typically, the transfer functions map ranges of densities and gradient values to the
