[Radiance-general] Testing rpict's -aa 0 option

[Georg Mischler]

Mark Stock wrote:

> Hello,
>
> After learning quite a bit about rpict's "other" options
> while working on my image for SIGGRAPH, I decided to run
> some tests to determine optimal settings for -ab, -ad
> paramters, and required oversampling to achieve smooth
> ambient results on surfaces, and smooth penumbras, given
> the -aa 0 setting for ambient accuracy.
>
> The page would be complete save for a few more renderings
> that are in the works right now. Here is the link:
>
> http://mark.technolope.org/radiance/aa0_test/final.html
> (59 images, ~2 MB total)
>
> My conclusion is that penumbras require high oversampling,
> and at those levels, a very low number of ambient divisions
> are required for a smooth rendering (-ad 4 to 16).
>
> I welcome any comments that you may have on the method,
> results, etc. I may have forgotten some very important
> parameter that throws all of these results askew---if I
> did, please tell me! =0)

Hi Mark,

it's been a while that you asked this, but I just ran into
your mesage when cleaning out my mailbox.

I think the main problem with your experiments is in the default
parameter values of rpict that you mention at the top of the
page. As mentioned in the rpict man page, when using -ds and -dj
for getting penumbras, you need to set -ps to 1 or you'll get
very uneven results. Rendering at multiplied resolutions can
compensate only for a part of the pixels that are skipped in the
calculation with -ps 4.

While you ended up needing up to 8x and more resolution, I think
you'd get acceptable results at no more than 3x with -ps 1.


-schorsch

-- 
Georg Mischler  --  simulations developer  --  schorsch at schorsch com
+schorsch.com+  --  lighting design tools  --  http://www.schorsch.com/