|
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hello all,
as I am not so up-to-date with the cutting edge of pov ray
developement, I have a question. My apologies if it is a FAQ.
I am looking for an enhacement for the normal maps, that adds
small real displacements to the intersection point in the same way,
as the normal map perturbes the normal vector.
Doing real/unlimited deformations to arbitrary objects is difficult
and it seems that you have to revert to generate a mesh form the
object and deform the mesh.
The obvious other aproach -- of course -- is to use ISOs and encode
additional surface irregularities into the function defining the surface.
But I am looking rather for a solution that keeps the distinction
between modeling and texturing. E.g modeling the bricks of a wall
and then tossing in some dust, making it bumpy or rough here and
there and the like.
It seemes possible to exploit the fact, that such displacements
would be small compared to the size of the object. It would be
sufficient if there is some displacement that roughly follows
the structure of the pattern, that the surface as a whole is
kept intact and that shadow calculations are performed in a
similar manner so that the shadows don't look obviously wrong.
No need to be mathematicaly correct with respect to the
displacements though.
The background why I'm intrerested in such a feature is: I am
using pov-ray mainly as a means to do stereoscopic imaging. Wheras
the trick to perturb the normal works great in 2d images, it is of
rather limited use for stereoscopic images. An example would be a
water surface with some ripples. When traced stereoscopicaly with
sufficient resulution (to get the problems with AA down), it is
imediately obvious to the viewer, that the surface you see is
completely flat. The fact, that there are lighting differencies
introduced by the normal map makes it rather worse. This is because it
addes more texturing to a given point, and thus enabeles the human
3d-viewing system to decode the actual depth of the point more clearly.
My question is: does someone know of experiments, patches, custom
versions etc. that go in the described direction. I.e. "adding local
small surface perturbations governed by procedural textures".
This is not a POV-Ray 4 feature request, as I am able to add the
features I need to povray myself and just try if it works. I rather
ask because I do not want to reinvent the wheel.
Hermann Vosseler
Post a reply to this message
|
|
| |
| |
|
|
From: Christoph Hormann
Subject: Re: displacement instead of normal map
Date: 21 Feb 2002 15:01:48
Message: <3C75522A.AD4F35B4@gmx.de>
|
|
|
| |
| |
|
|
>
> [...]
>
> But I am looking rather for a solution that keeps the distinction
> between modeling and texturing. E.g modeling the bricks of a wall
> and then tossing in some dust, making it bumpy or rough here and
> there and the like.
In short: this won't be possible. Trying to make real geometric
displacement in a raytracer without a mathematically consistent method
will produce rubbish. There are methods to generate displacement mapped
meshes with high resolution 'on the fly' without storing the whole mesh in
memory, but this won't have any advantages speed-wise.
Christoph
--
POV-Ray tutorials, IsoWood include,
TransSkin and more: http://www.tu-bs.de/~y0013390/
Last updated 06 Feb. 2002 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
This just came to mind, and may or may not work... try a transparent plane
above the plane which you want to be distorted, and give the transparent
plane a bumpy normal, and then give it refraction. That might possibly
create a useable effect. Maybe not though.
- Slime
[ http://www.slimeland.com/ ]
[ http://www.slimeland.com/images/ ]
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> In short: this won't be possible. Trying to make real geometric
> displacement in a raytracer without a mathematically consistent method
> will produce rubbish. There are methods to generate displacement mapped
> meshes with high resolution 'on the fly' without storing the whole mesh in
> memory, but this won't have any advantages speed-wise.
Is this how BMRT does it?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>>But I am looking rather for a solution that keeps the distinction
>>between modeling and texturing. E.g modeling the bricks of a wall
>>and then tossing in some dust, making it bumpy or rough here and
>>there and the like.
Christoph Hormann wrote:
>
> In short: this won't be possible. Trying to make real geometric
> displacement in a raytracer without a mathematically consistent method
> will produce rubbish.
Quite true.
Of course such a method has to work consistent, as we can make
in general almost no asumtions about the geometry of the object a
given intersection belongs to.
Hermann
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Slime wrote:
> This just came to mind, and may or may not work... try a transparent plane
> above the plane which you want to be distorted, and give the transparent
> plane a bumpy normal, and then give it refraction. That might possibly
> create a useable effect. Maybe not though.
yes, this works.
But I fond myself bound to use several stacked planes with maps
adjusted acordingly. Not very easy to handle.
Hermann
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |