There are, roughly speaking, two possible approaches to raytracing:

1) Use the basic backwards raytracing algorithms and simple lighting
   models (essentially the phong lighting model, perhaps some other
   similarly simple lighting models as well).

2) Try to raytrace *everything*, use full BRDFs for surface definitions.

  The advantage of method 2 is that, assuming you have good BRDFs, you
can get quite impressive results at times, while in method 1 there's a
limit on how realistic the image can be.

  The disadvantage of method 2 is that it's not only very slow, it's
*always* slow. You simply *can't* create a very simple scene (eg. a
sphere on a plane) and render it in 0.5 seconds, like you can with
method 1. In many cases the latter is the more desirable method for
this precise reason: It often gives acceptable quality very fast.

  Trying to support both methods at the same time is a bit problematic
too. It's hard to mix them. Basically you would need two separate
rendering engines and a way to choose which one is used where. The
full BRDF definitions are basically useless in method 1.

  If you want POV-Ray to support raytracing with BRDFs, that would
mean basically that two different raytracers would need to be packaged
into one executable, and you simply choose which one is used. The advantage
in this can be dubious. It may be much better to simply have two different
programs.

-- 
                                                          - Warp

Post a reply to this message