>       count 10
>       nearest_count 5
>       error_bound 1.8

These values alone could cause quite some problems this way. To sample the
lighting, you only shoot 10 rays into the scene per radiosity sample. You'll
miss large sections of the scene that way, and get very imprecise sampling
results.
error_bound is set too high, so it's quite possible that when averaging
neighbours of samples (nearest_count will search for 5 radiosity samples in
close proximity and average the results for a smoother lighting), you take
some of the very bright ones inside the door into account somewhere.
Additionally, the sampling with only 10 rays could be the source for the
bright scenes inbetween, when more of these rays hit the bright door instead
of some other area.

Finally, all you're objects seem to be glowing if you use ambient 0.1 on all
objects for the radiosity. You should only have lightsources (which, in the
case of radiosity, doesn't have to be the literal light_source, but may be a
large object with a high ambient value) with high ambience values. The
sky_sphere also works like a lightsource, which is something to keep in mind
for outdoor-scenes.

Regards,
Tim

-- 
"Tim Nikias v2.0"
Homepage: <http://www.nolights.de>

Post a reply to this message