 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hiya.
I've had a root around but can't find any help on this problem.
I'm trying to make an image with caustics cast by Bezier patches.
With photon spacing 0.01 I get suspiciously blurry caustics, so I tried photon
spacing 0.001 and the resulting caustics are made up of conspicuous flat-shaded
quadrilaterals.
You can see the 2 pics at:
http://viewmorepics.myspace.com/index.cfm?fuseaction=user.editAlbumPhoto&albumID=344068&imageID=54508714
and
http://viewmorepics.myspace.com/index.cfm?fuseaction=user.editAlbumPhoto&albumID=344068&imageID=54508715
(Actually, having seen the 0.001 pic it's easy to spot the quads in the 0.01
pic. They're there but blurry.)
Any suggestions on photon settings I could try to make the caustics sharp and
realistic? Or is it a problem with the Bezier patches?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Bruce Mardle" <mar### [at] yahoo couk> wrote:
> You can see the 2 pics at:
http://c1.ac-images.myspacecdn.com/images02/84/l_ca60ffd5b7874015beeb13480dd9baa8.png
and
http://c1.ac-images.myspacecdn.com/images02/110/l_9fbf233e1ad04ec8879ba1e4744c8e30.png
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Bruce Mardle schrieb:
> Any suggestions on photon settings I could try to make the caustics sharp and
> realistic? Or is it a problem with the Bezier patches?
POV-Ray actually works with a mesh-like pproximations of the patch, not
with the exact mathematical representation. You can increase the
resolution of that mesh using the following parameters on the bezier
patches:
type 1
u_steps 3
v_steps 3
flatness 0.0
The first parameter tells POV-Ray that you want control over the mesh
resolution, at the cost of some memory (type 0 would be default).
The second and third parameter tells POV-Ray the (maximum) mesh
resolution you want. Note that the actual resolution is 2^u_steps by
2^v_steps. Increase this to reduce the artifacts you observe, or
decrease for speed.
The fourth parameter tells POV-Ray at which curvature to stop
subdividing even before reaching the maximum; a value of 0.0 tells
POV-Ray to always subdivide to maximum. This can speed up things, but
may introduce other artifacts ("cracks" in the patch).
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thanks, clipka.
Actually, I'm using Moray, which, as far as I can tell, always outputs bicubic
patches with "type=1". Fortunately, I can tweak the .POV file on its way from
Moray to POV-Ray.
Moray's default for the patches is type=1, u_steps=v_steps=3, flatness=0.01. I
get much better, and near-identical, results using either type=1,
u_steps=v_steps=5, flatness=0 or type=0. (The latter's somewhat slower.) Photon
spacing=0.01, either way.
If I use photon spacing=0.001, the caustics get sharper, as expected, but I also
get a fairly subtle 'ripple'. It'd be nice to get rid of that, but I can live
with it. (Curiously, the output is equally ripply with photon spacing=0.002.)
Spacing=0.01:
http://c2.ac-images.myspacecdn.com/images02/88/l_d0210a72b0f24dbdbeb58c3d6f9f5bf5.png
Spacing=0.001:
http://c3.ac-images.myspacecdn.com/images02/104/l_2c4db7de4cf24bfa8df38a41ab2b0352.png
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Bruce Mardle schrieb:
> If I use photon spacing=0.001, the caustics get sharper, as expected, but I also
> get a fairly subtle 'ripple'. It'd be nice to get rid of that, but I can live
> with it. (Curiously, the output is equally ripply with photon spacing=0.002.)
It appears to me that this is your bicubic patch artifacts showing once
again; in that case I'd expect the "ripples" structure to be quite
independent of photon spacing, except when the photon spacing is high
enough to blur it.
Maybe you want to choose a higher photon spacing for the bezier patch
than for the sphere, using "photons { target 10 ... }" (for instance;
this will give a 10-fold higher photon spacing than specified in the
globals section).
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thanks again, clipka.
I recently, unintentionally wiped by Windows XP machine, so it'll be a few days
before I try your suggestion.
What I'm ultimately hoping to do is simulate ripples moving on water and make a
ray-traced animation of it. I might be better off using a heighfield, rather
than bicubics.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Bruce Mardle wrote:
>
> What I'm ultimately hoping to do is simulate ripples moving on water and make a
> ray-traced animation of it. I might be better off using a heighfield, rather
> than bicubics.
>
>
>
simulation. Tim Nikias created a set of macros that can be downloaded here.
http://www.nolights.de/downloads.html#lssm
--
Best Regards,
Stephen
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thanks, Stephen.
I've got a "virtual ripple tank" but your suggestion sounds like it might be
less work!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Bruce Mardle wrote:
> Thanks, Stephen.
> I've got a "virtual ripple tank" but your suggestion sounds like it might be
> less work!
>
>
>
I tested it when Tim first developed it and it worked quite well. Well
enough for me to abandon my own efforts :)
--
Best Regards,
Stephen
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
