Physics and Depiction
One sometimes hears or reads (variants of) the following:
"The physical quantity that affects the eye is intensity. Models of
realistic depiction should therefore be based on the theory of radiative
transfer."
Some comments:
Radiative Transfer.
Phenomenological theory for the propagation of light. Light is modelled as
a physical quantity having the units of power per area per solid angle.
An integro-differential equation for this quantity can be derived from simple
phenomenological considerations. This equation along with boundary
conditions can be solved in principle to compute the radiance.
at each point of space and for all solid angles.
If the theory is used to produce pictures, than it should also incorporate
a model of how the a light source is reflected from the picture. Someone
sitting in a bathtub surrounded by candles will perceive the SIGGRAPH
proceedings differently than someone sitting on a camel in the middle of
the desert.
Actually people interested in reproducing pieces of art on electronic media
have paid attention to these "details".
It only makes sense to talk about the eye really if
the light impinging on the eye would be simulated directly. Building the
hardware that would achieve this seems to me to be a formidable engineering
task. Not to mention the dangers associated with such a piece of
hardware.
There is nothing fundamental about the theory of radiative transfer, it
is after all a phenomenological theory. There have been many attempts to
relate it to Maxwell's Wave equations. It is commonly assumed that the
relation is given by Poynting's vector. As Wolf has shown the relation is
actually far more subtle than that, and is related to the state of
coherence of the wave. Perhaps, Poynting's vector is good enough for
the limit of small wavelength and assuming that we are far enough away
from the light source (and that the source is incoherent). Are these
reasonable assumptions in computer graphics ?
Why stop at Maxwell's Equations ? Isn't the quantum theory more fundamental ?