[Bf-committers] Re:Improving the quality of soft shadows

Alfredo de Greef bf-committers@blender.org
Wed, 17 Dec 2003 15:49:34 -0800 (PST)

> I remembered a paper a read a year or so ago that
> had to do with a new sampling
> stratagey, called "interleaved sampling".
>   I think that it could be applied to the
> soft-shadow sampling, and I suspect
> that doing so would improve the quality of the soft
> shadows considerably
> without much expense.
> (It could also, perhaps, even be incorperated
> into other over-sampling based
> features in Blender such as anti-aliasing and
> motion-blur).

Funnily enough (or not..) I have actually done this,
but I'm not so sure it would really be considered an
improvement as far as shadows is concerned. The
dithered look (like ordered dithering) can be quite
obvious. I used a Halton sequence as irregular basis.
Before that I also tried randomized QMC, again using
the Halton sequence, but although that has the option
for an early out which makes it faster (after say 16
samples of 64 total you can check if the shadow
samples are all shadowed or unshadowed since Halton is
independent of sample number), it can be a bit noisy.
So probably an adaptive approach would be best. The
interleaved scheme might work very well for AA and
motion blur though as demonstrated in the paper.


Do you Yahoo!?
Free Pop-Up Blocker - Get it now