Graphics Reference
In-Depth Information
state node. Also stored are the SVD matrices
U
u
and
U
a
, the columns of which
are displacement and transfer basis vectors, respectively. It is assumed that an
arbitrary deformation can be approximated by the linear combination of IRFs, so
at run time there is nothing left to do but synthesize the shape and the appear-
ance for the particular lighting condition and deformation state. This can be done
efficiently in graphics hardware.
The method described in the “deformable scenes” paper became a hot topic
in both research fields—dynamics and rendering—soon after it was published.
It deftly combines the results of the precomputation of dynamic simulation and
global illumination. Of course using this kind of union limits the amount of an-
imation that can be performed. Regardless, research to expand radiance transfer
to animation increased dramatically as a result of this paper.
10.4 Precomputed Acoustic Transfer
Another area that was influenced, rather unexpectedly, by the PRT work of Sloan
et al. was the field of synthetic acoustics, also known as
sound rendering
. The goal
of sound rendering is to procedurally synthesize sound in synchronization with
computer-generated animation. The paper entitled “Sound Rendering” by Tapio
Takala and James Hahn presented at SIGGRAPH is considered to be the genesis
of this particular subfield of computer graphics [Takala and Hahn 92]. Mirroring
the history of rendering in general, physical accuracy was not initially empha-
sized; rather, physically based approaches were incrementally added. Sound ren-
dering consists of two processes: sound generation and sound propagation. The
research that enabled physically accurate sound generation appeared in the 1990s
(e.g., [van den Doel and Pai 96, O'Brien et al. 01, van den Doel et al. 01, Dobashi
et al. 03a]). Research related to sound propagation in CG environments had been
in progress since the late 1990s, but computing accurate sound radiation from
geometrically complex vibrating objects was typically ignored. Diffraction and
interreflection effects in an environment are an important part of accurate acous-
tic modeling, and these were largely ignored until precomputed acoustic transfer
(PAT) emerged.
Sound effects caused by global sound radiation are analogous to global illu-
mination effects. Sound and light are both wave phenomena and there are sim-
ilarities in how their propagation can be simulated; for example, ray tracing in
acoustics predates ray tracing for CG rendering by many years. An important dif-
ference, though, is that wave effects such as diffraction and interference are typ-
ically unimportant in illumination, whereas they can be very important in sound
modeling. Doug James, who had connected physically based deformation with
