Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
all positions, directions~ and energies. This is a notoriously the behavior of large numbers of particles obeying the pre-
difficult equation to solve analytically in all but the most scribed laws of motion expressed as scattering probabilities.
trivial problem instances [2,5]. This is true of the rendering Each particle undergoes a sequence of collisions or scatter-
equation as well which is essentially a variant of the linear ing events which probabilistically alter its trajectory at each
Boltzmaun equation. The principal difference is t h a t the collision-site and contribute to the history of the particle.
scattering kernel is rephrased as a geometry term, g~ which Each particle history, or random walk, is used as a statistical
accounts for occlusion and inverse square attenuation, and estimator of average case behavior. R a y tracing is a mecha-
a trivariate scattering term, p, whose arguments are sur- nism for computing points of collision, and a stochastic ray
face points (See [10]). The latter encodes the directions of p a t h [10] is the resulting r a n d o m walk of a particle. The
incidence and reflection implicitly through the positions of rendering equation provides a link which allows us to view
the source and destination elements relative to the point of image synthesis in terms of particle transport. T h r o u g h this
reflection. The similarity of the rendering equation to the connection we can gain useful insight into the features and
linear Boltzmann equation suggests that m a n y of the power- limitations of image synthesis techniques.
ful techniques which have been developed for other particle For example, consider the use of decoupled passes of ray
transport problems may be applied to problems in image tracing and radiosity to model specular and diffuse modes
synthesis. of transport independently. It has been observed that sim-
We note t h a t there are several aspects in which the ren- ply combining the results of these passes fails to account
dering equation is somewhat more tractable t h a n the trans- for some important phenomena of geometrical optics [16].
port equations in fields such as nuclear engineering. The most obvious example is a caustic formed by specularly
transmitted or reflected light falling on a diffuse surface.
1) The particles (i.e. photons) do not influence one another, B o t h classical ray tracing and radiosity totally neglect this
alter the environment, carry a charge, or replicate via mode o£ transport, therefore this deficiency cannot be reme-
fission. Thus scattering is independent of • as well as died by summing their contributions a posteriori. Wallace
external forces, making the equation linear. describes a solution for this particular case of specular-to-
diffuse transport, but it is impossible to account for all such
2) In the absence of participating media, collisions occur
sequences of transport as special cases. This phenomenon
only at surfaces. The particles therefore have a rela-
has been observed in other linear transport problems and
tively large mean free path.
is attributed to the fact t h a t equation 1, t h o u g h linear in
3) We seek only the steady state solution, not transient the source term, S, is nonlinear with respect to the scat-
distributions on the way to equilibrium. tering kernel, K . While the linearity in S allows us to sum
the independent contributions made by different sources and
These properties manifest themselves largely in the rela- wavelengths of light, the analogous decoupling fails when the
tively simple form of the scattering kernel which is com- kernel is partitioned into, for example, Kspec+Kdiff. A faith-
prised of the bidirectional reflectance functions associated fnl simulation of all modes of transport can only be achieved
with the surfaces. After probabilistically determining a new by coupling t h e m in the solution process.
particle direction at each scattering event the next col].ision-
site along the r a n d o m walk is completely determined, elim- 3 Russian Roulette
inating stochastic distance calculations. However, there are
two respects in which this transport process is made more The albedo of a surface is the probability that an incident
d i ~ c u l t t h a n typically encountered in other disciplines. particle will be re-radiated after collision [3]. In Monte Carlo
First, the geometry of the simulated environments can simulations this probability is normally used to adjust a nu-
be arbitrarily complex. While simulations of reactor cores merical weight associated with the particle rather t h a n prob-
and semiconductor devices benefit from fairly constrained abilistically terminating the history. This technique, termed
geometries and exploit special properties of lattices, cylin- implicit capture [12], has better statistical properties owing
ders, slabs, etc. [12], the trend in computer graphics is to to longer particle histories.
move toward greater and greater scene complexity. This A p r o p e r t y of implicit capture is t h a t particle histories
is exemplified by recent work involving billions of geomet- can only terminate at surfaces of zero albedo or by leakage,
rical primitives [14]. This can be further complicated by t h a t is, by escaping the system. However, it is nearly always
time-dependent scene geometry. Simulation of the resulting impractical to continue tracing a p a t h until one of these
m o t i o n blur requires time averaging steady-state solutions conditions is met. Even if we could guarantee the eventual
at intermediate scene configurations. termination of every history, we would spend an inordinate
Secondly, the problem of interest in image synthesis is to amount of time computing collisions involving particles of
compute the intensity of illumination impinging on a single negligible weight. One solution is to place a limit on the
point, the "eye", through small apertures which correspond number of scattering events in a particle history and to ig-
to "pixels". Analogous situations occur in reactor shield- nore all contributions beyond this point. A better solution
ing problems which simulate point radiation detectors [3]. is to use weight cuto~ which truncates the particle's history
These are inherently more diiBcnlt to solve than the typical only when its weight falls below some threshold [12]. The
problems which involve flux averages over volumes. idea of using weight cutoff to terminate ray tracing recursion
M a n y important problems in particle transport do not was introduced by Hall [8] and termed adaptive tree depth
admit analytic solutions and are also prohibitively expensive control. B o t h of these techniques are commonly employed
to solve via numerical integration due to the high dimension in ray tracing implementations in order to avoid excessively
of the phase space in which they operate (e.g. three spa- deep ray trees and, in extreme cases, even unending recur-
tial dimensions, two directional dimensions, and an energy sion due to opposing mirrors or total internal reflection. The
dimension). The only recourse for solving these types of difficulty with this type of policy is that truncation intro-
problems appears to be Monte Carlo methods which track
64
~ ComputerGraphics,Volume24, Number4, August 1990
65
O StGGRAPH '90, Dallas, August 6-10, 1990
65