Greg Edwards <edw### [at] hotmailcomremovethis> wrote:
> A much better solution which I was thinking of for a while would be to 
> scrap the radiosity system and do radiosity calculations using photon 
> mapping. Photon mapping was originally meant to be a full global 
> illumination model afterall.

  Old idea, and AFAIK Nathan has tried it with not too good success.

  If you think about it, global lighting with photon mapping would be
the exact reverse of the current stochastic global lighting in POV-Ray
(ie. the photon mapping method sends rays from the light sources to the
scene, while the latter method sends rays from the current surface point
to its surroundings).
  There are advantages in the latter methods, which I can't find in the
photon mapping method. Photon mapping is excellent for reflections and
caustics because the rays can be shot to a rather small portion of the
scene (more precisely, towards the bounding box of the reflecting/refracting
object), but if you want to calculate the lighting of the whole scene,
you'll have to shoot rays to the whole scene, which means pretty many rays.
Many of these rays will be shot in vain (because they will end up in parts
of the scene not seen from the camera).

  The stochastic global illumination method used by POV-Ray is not problematic
per se, it's actually a good one and used by other renderers as well (eg.
Radiance). The reason for the problems in this particular implementation are
not clear. It might be buggy or not well implemented. I don't know.

-- 
#macro N(D)#if(D>99)cylinder{M()#local D=div(D,104);M().5,2pigment{rgb M()}}
N(D)#end#end#macro M()<mod(D,13)-6mod(div(D,13)8)-3,10>#end blob{
N(11117333955)N(4254934330)N(3900569407)N(7382340)N(3358)N(970)}//  - Warp -

Post a reply to this message