[Bf-committers] Proposed Method for GI in blender internal

[Anders Nilsson]

>From a look on your proposal it looks exactly as photon mapping except
the irradiance is stored in lightmaps instead of in the photon map. This
require splitting the scene into patches, just as for radiosity. This
has been done with photon mapping before, Jensen mentions
implementations of it in his paper (illumination maps) although he
chooses not to use it since the photon map can be used for any geometry
(perfect spheres and fractal strucures as well as polygonal objects).

One problem is that caustics needs very small patches in order to look
good. The indirect diffuse reflection works ok since you do a final
gather pass to get them (you just have to either way).

I don't think this would be much faster than photon mapping, perhaps
photon mapping could be speeded up by precomputated irradiance stored in
each photon (I find it 10 times faster and somewhat uglier) or something
else but I still think photon mapping is the way to go here. Smoothing
it would not be good enough, final gather is still needed.

Once the photon map is build it can be used to calculate say
illumination-per-patch for your method but the problem with caustic
resolution still persists.

Then on the other hand I could be totally wrong so by all means try it
:) (and pardon me if I misunderstood your description of the algorithm)

Anders Nilsson (breakin)


On Fri, 2004-06-04 at 23:15, Ray Wells wrote:
> I came up with a cool idea for quick GI in blender
> internal..
> 
> 
> if anyone wants to give me a hand as far as
> implementation that'd be really cool.
> 
> http://vivi2.no-ip.info/quickgi.html