Thank you Mark, for your very clear explanation! Arjo. -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of Mark Heuymans Sent: dinsdag 13 maart 2007 23:02 To: [email protected] Subject: Re: groundfog test project
Arjo,
Where's the chapter about meditation in the RS manual? I need it ;)
What I got so far:
- volume shadows without any form of randomization leads to banding
artifacts. In fact, I noticed that volume sampling of 200 produces pretty
clean results, but that's a bit too much (however: skipping the
randomization code makes the shader faster)
- randomization of the Coordinates (a different channel from Map coords!)
partly reduces banding but introduces grain. The problem is only partly
solved because the Map coords don't match position (Coordinates).
- the Map object takes care of this by re-computing Map coords from the
new randomized position. Turbidity and volume illumination are calculated
from identical positions inside the volume; artifacts are greatly reduced.
In this case, the Random function is used rather than a Noise object: it's
something completely different. Noise is a fractal pattern generator, Random
doesn't have any patterns or scale/frequency.
Nice tiles btw; in your example Noise is obviously the way to go, rather
than Random!
my 2 cts (I doubt it helps... hit me with a hammer if I'm missing the
point)
-Mark H
This is what I ment for the randomize coords:
The code...
in the result it slightly randomizes the picture that has been applied
to the object with a parallel mapping object. (I used the floortiles that
are in the texture folder)
I'm proabably completely off track with this approach. I haven't
explored volume shaders much so far. So maybe it's better to start doing
that first and then it might get clear by itself.
The thing I don't understand is what you mean with "position". What
position are you referring to?
Arjo.
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] Behalf Of Vesa Meskanen
Sent: maandag 12 maart 2007 17:47
To: [email protected]
Subject: Re: groundfog test project
Hi,
> I think I understand it. But if I understand it right then I don't
> understand why you don't put the randomization before the turbity
object.
It is just a matter of the order of the computations. When you
randomize the position, you need to compute the new map coords value from
the new position. It was earlier done by the material map object (parallel,
sphere or whatever map - the material should actually be independent of
that). When the position changes, how can you re-evaluate Map coords again ?
How does the VSL code know what kind of parent mapping was applied (what
type, what size etc?). The answer is: using the Map vsl object, which
re-evaluates the parent mapping.
> Ray tracer: render engine computes position
> Material mapping: Parallel map computes Map coords from
position
> Material:
> randomize map coords **** how ?????
> Position changes by randomization
> Turbidity is computed from randomized Map
coords
I highlighted the important row above. How can you randomize the map
coords channel so that it's randomization matches the randomization oif the
position. Maybe you Map coords randomization goes to negative x axis
direction, but position goes to +x axis direction! I repeat: Map coords
depends on the size, orientation and type of the parent mapping, and VSL
code normally does not know these.
Map object takes care of such details:)
Best regards,
Vesa
<<attachment: code.jpg>>
