Ahead of everything else: I'm asking
stuff, so to all those "newbies" to radiosity:
don't expect this to be correct!

...

Okay, I've read through the paper, but I'm
no mathematician and haven't fully understood
all details. This is what I think POV's radiosity
does, I'd be happy if anyone could correct me:

First, it shoots rays outwards. The position where
it hits will then send "count"-many rays outwards,
in an even distribution, to collect light-values from
"outside". This is the Monte-Carlo-Part of the
algorithm.
Then, as a new ray is shot, it tests if other samples
are nearby, and if there aren't enough (set by varios
details like minimum_reuse, error_bound etc), I
place a new sample there.

But what about recursion_limit? As I understand it,
I do that all at once: shoot rays outward from first
sample. The places where these hit will also shoot
samples outward, and so, until recursion-limit is
met. This way, I quickly get samples across the
scene in the very beginning, which may then be
accessed for the next samples. For this to work,
the samples would have to get marked how deep
they've already done their own recursion_limit... Tricky.
Perhaps someone could explain that to me?

And finally, pretrace. As I understand it, pretrace is
just another pass over the picture in order to check
if enough samples are present at any given spot. If
I'd be crazy enough, I could test this for every pixel
of a scene (though this doesn't necessarily lead to
best results, I've understood that much at least :-)

I'm really curious how all this works together. I formerly
thought that POV-Ray creates patches across the
scene, but that doesn't seem to be the case. Instead,
it averages samples points in a given radius for a pixel.
Which leads me to the conclusion that the radiosity
can't "detect" shadowlines, and do better sampling
there. Right?


-- 
Tim Nikias v2.0
Homepage: http://www.digitaltwilight.de/no_lights

"Tim Nikias v2.0" <tim### [at] gmxde> schrieb im Newsbeitrag
news:3f0f50b5@news.povray.org...
> I hope I'm not in the wrong group when asking
> the following question here, if I am, please forget
> that you know me. :-)
>
> How does POV-Ray's Radiosity method work?
> I've tried searching the web for a document covering
> the actual function and approach, but was unable
> to find something useful.
>
> Anyways, as far as I've understood the docs, its
> like this:
> 1. Pretrace-Level:
> Divide scene into patches, calculate ambient light-values
> for those, gathering diffuse-reflected light by shooting
> rays (amount given by "count") outward.
>
> For each successive step, we'll check if neighbouring
> patches are too different in brightness, and subdivide
> them. If we use a recursion_limit, we'll do another
> "bounce" of light outward, though I'm unsure if that
> happens in a new pretrace-step, or is done right at the
> beginning, and further pretrace-steps just do the
> subdivision.
>
> That's just about how far I was able to understand
> it, but I'd like to know it in detail. For example, if
> someone could point me on a document describing
> the technique, and provide some information on what
> POV-Ray has used and what was modified...
>
> The main thing I'm actually trying to do with this question
> is gather information on what different pretrace-steps do,
> if and which some calculations are done iterative (and thus
> the need for several pretrace-passes), etc. I'd like to really
> fully understand the individual options for radiosity, cause for
> some I've just got the feeling like:
> use lower values for more detail (regarding pretrace),
> use higher values for more accuracy (regarding count),
> etc.
>
> Regards,
> Tim
>
>
> -- 
> Tim Nikias v2.0
> Homepage: http://www.digitaltwilight.de/no_lights
>
>
>
> ---
> Outgoing mail is certified Virus Free.
> Checked by AVG anti-virus system (http://www.grisoft.com).
> Version: 6.0.495 / Virus Database: 294 - Release Date: 01.07.2003
>
>


---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.495 / Virus Database: 294 - Release Date: 30.06.2003

Post a reply to this message