Graphics Reference
In-Depth Information
tegral. Monte Carlo ray tracing is a form of Monte Carlo integration: the random
points in the domain of integration are the random ray directions. Monte Carlo
methods are sometimes called
stochastic
methods.
2.3.2 Basic MCRT
The first ray-tracing implementation using a Monte Carlo method was described
in a SIGGRAPH paper entitled “Distributed Ray Tracing” by Rob Cook, Thomas
Porter, and Loren Carpenter [Cook et al. 84]. Distributed ray tracing makes effec-
tive use of
supersampling
, the casting of many rays at each pixel and from each
intersection point
(
Figure 2.4(a)
)
. Glossy reflection at a surface point is computed
by averaging multiple reflected rays according to the spread
distribution
of the
specular reflection. In the original paper, the implied BRDF is rather simple: the
reflected rays are sampled uniformly in a narrow cone about the direction of mir-
ror reflection as if the BRDF has the shape of the cone. The width of the cone thus
determines the specularity. The implementation also uses Monte Carlo sampling
of area light sources to generate “soft shadows”
(
Figure 2.4(b)
)
.
Specular
Viewpoint
(a)
Light source
Diffuse
Viewpoint
(b)
Figure 2.4
Distributed ray tracing averages the results of a collection of sample rays. (a) Reflection is
computed by tracing a bundle of rays distributed according to the BRDF; (b) tracing rays
to enough points distributed across an area light source produces soft shadows.
