What are highlights?  They are specular reflections, often blurred, of
light sources.  So it dawned on me circa 2013 August 25 that if my
objects have both reflection and specular/phong in their finish, they
shouldn't also be reflecting the Sun or light bulb.  No prob, just set
no_reflection on the looks_like for the time being.  For my desk lamp
project, my idea has been to have a flag for whether or not the light
bulb should reflect, thereby allowing the user to choose between
highlights and a reflected bulb.

Enter radiosity artifacts from lamp fixtures.  (See "Hooded lamp
occlusion" in p.b.i.)  The inner surface of a lamp hood is so bright
that radiosity leaves bright spots all over the scene unless an insanely
high radiosity count is used.  Setting no_radiosity on the hood interior
eliminates the artifacts, but now the light from that radiosity must be
replaced.  A 180 degree cosine falloff spotlight does the trick.

But now, this means that objects with both reflection and highlights
will now reflect double the light from the hood interior!  I cannot set
no_reflection on the hood interior, because then the texture of the hood
exterior would show through in the reflection.  Besides, the hood is
larger than the bulb, and the more extended an object, the bigger the
morphological discrepancy between its reflection and the equivalent
point highlight.  What was easy for the light bulb doesn't work for the
fixture.

So how should this dilemma be resolved?  Have the user refrain from
setting specular or phong in finishes that have a reflection?  Just live
with the double light, since it would most often exceed the image
format's dynamic range anyway?  The former solution would require high
quality (i.e, slow) blurred reflection for realistic highlights, and
latter solution would be unsuitable for HDR and EXR images.  Is there a
third way that doesn't involve the user in these complications?

Or should I just go ahead with the reflection flag, and warn the user
that the flag won't affect the hood interior?

Post a reply to this message