>> This is what happens IRL though, isn't it? An sRGB red light source will
>> have a very narrow spectrum and almost no overlap at all with the
>> reflective spectrum of an sRGB blue surface, so it will appear black.
>
> OTOH if you have eg. a pure yellow light, it will look to the human eye
> the same as a light with red and green frequencies in appropriate
> proportions, but they will illuminate surfaces in radically different
> ways. (White surfaces will look about the same under both lights, but
> eg. red surfaces won't.)

The difference with yellow (or any colour other than pure red, green or 
blue) in the current RGB system is that it needs at least two non-zero 
components (R and G in this case). This means it will illuminate any 
surface with a non-zero component in R *or* G in this case. The effect 
of this is to simulate a relatively broadband spectrum - only surfaces 
with R *and* G zero (ie a pure blue) would appear black.

Obviously this is a limitation (that no colour other than the sRGB 
primaries can be simulated as a narrow-band spectrum) - which is what I 
meant when I suggested that spectral rendering would improves scenes 
with narrowband lights (or even lights with an "interesting" spectrum).

Post a reply to this message