Graphics Reference
In-Depth Information
Figure 8.13
Images rendered using the Ward model. (From [Ward 92] c
1992 ACM, Inc. Included
here by permission.)
interpolation problem. The dense sampling of Ward's image-based measurement
system eliminates the need for interpolation in the outgoing direction, and greatly
simplifies it over incoming directions. A drawback, however, is that the measure-
ment precision is limited by the dynamic range of the camera, and even further
by the fact that the measured values come from dividing by the reflectance of the
reference sample, which is measured separately. Spectral sensitivity is also an
issue, which in Ward's arrangement is limited to the three-channel output of the
digitized camera image. In contrast, sensors are available that have a much higher
dynamic range and greater spectral sensitivity.
The original goal of the work described in Ward's paper “Measuring and Mod-
eling Anisotropic Reflection” was the construction of the measurement system.
However, he felt that a paper describing the measurement method alone would be
less useful without an accompanying BRDF model, so he formulated the model
a few weeks before the submission deadline. Ironically, the greatest impact of
the paper turned out to be the BRDF model rather than the measurement sys-
tem. Ward himself never imagined the model would end up in such widespread
use.
8.1.6 Generalization of BRDF Models
Ward's BRDF model was unusual in that it was developed for the specific purpose
of fitting measured reflection; most other models were based either on a simple
heuristic like the Phong model, or derived from assumptions about the specific
surface, i.e., microfacet models. In the 1990s, new developments for accurately
measuring the BRDFs of real surfaces opened up possibilities for a different ap-
proach. Fitting parametric representations to the measured data made it possible
to represent various types of reflection using the same basic expressions.
