RE: [Algorithms] Static Lighting using Compressed Light Probes

[Jon W. <hp...@mi...>]

> > Occusion "factor" is just percentage of the unoccluded hemisphere.
> > Just shoot a bunch of rays and count unoccluded ones; no weighting
> > applied.

> I've experimented with just this ambient occlusion term.  It softens
shadows
> say at the base of the pillar, but doesn't give me the direct shadow of
the
> pillar.

I'm assuming the issue is that your light source moves? If so, then a
single ambient occlusion term would occlude as much on the ground IN
FRONT OF the pillar, as it would behind the pillar. Not good :-)

I think the basic problem is that you don't have enough precision in
your occlusion maps. Shadows can have fairly sharp edges, so for a crisp
line, you need many terms in your SH. An ambient term is like a 0th order
SH, so it only has a single term. However, even with a second order SH
(9 coefficients) you won't get "crisp" shadows, but at least you'll get
variation based on direction.

I e: for each texel, render the background as white. Then render the
nearby occluders in gray-to-black (darker the closer they are). Last,
turn this rendering into an SH using a SH best-fit techniques. In the
end, you have an occlusion SH that you can use to modulate incoming
light.

If you want crisper/better shadows, you can keep adding orders to the
SH (gets very expensive) or start using shadow maps or stencil shadows.

Cheers,

			/ h+