Am 11.04.2010 16:34, schrieb Tek:

> I get it in both Megapov 1.21&  Povray 3.6. I might try moving over to 3.7
> but my scene's probably using a couple of megapov features.

Um, sorry... I'm so focused on POV-Ray 3.7 that it didn't even cross my 
mind that you may still be using 3.6. >_<

>> What is strange, is the fact that the artefacts are of the complementary
>> colour.
>
> It's as if it's using negative lighting values.

It is indeed! Thanks for pointing me in the right direction, that did 
ring a bell:

At each radiosity sample location, POV-Ray 3.6 does not only compute the 
total incoming diffuse illumination, but also tries to guesstimate the 
illumination /gradient/ at that point.

I had always been suspicious of that mechanism's implementation, though 
I never managed to demonstrate its bogosity with a sample scene. On the 
other hand I couldn't make out any noticeable benefits either, so I 
decided to disable it in POV-Ray 3.7, if only to save memory and 
processing time.

To verify this theory, I just threw together a patched 3.7.0.beta.36 
with the "gradient code" re-enabled, and it did indeed show the very 
same complementary-color halos (current official 3.7 betas don't, as 
already mentioned in this thread).

Patching POV-Ray 3.6 (or MegaPOV, for that matter) to fix this issue 
should be pretty easy: In radiosity.cpp, just delete (or comment out) 
the line reading:

     #define RAD_GRADIENT 1

That should be all that's needed.

Post a reply to this message