|
 |
Nicolas George nous apporta ses lumieres en ce 22-04-2007 15:36:
> Chambers wrote in message <462ba23c@news.povray.org>:
>>> Neither solution are satisfactory. For a long time, I think that PoV is
>>> missing a way to cleanly deal with coincident surface: some declaration in
>>> the scene that says "these surfaces are *meant* to be coincident, you must
>>> handle them that way"
>> But how would it handle which texture to process first?
> That is exactly the point: we need a way to tell PoV what to do.
> From a theoretical point of view, it is quite trivial: if two objects use
> the same surface, PoV computes only once the intersection, applies the
> user-given rules to determine which one comes first, and keeps a flags for
> any light ray it must then follow from this point.
> From a practical point of view, devising a syntax for that would require to
> example all possibilities, but it can be done.
You have "normal" textures.
You have interior_texture. (an optional element)
Add "coincident_texture" (as another optional element) with a
coincidence_treshold parameter (defaulting to epsilon*2 for example) defining
the maximum gap that can separate the surfaces to be considered as coincident.
For those unfamiliar with the term, the "epsilon" value is the smalest value you
can have before it's considered as zero.
--
Alain
-------------------------------------------------
Christian Science: When shit happens, don't call a doctor - pray!
Post a reply to this message
|
|
|
|
Am Wed, 25 Apr 2007 18:49:09 -0400 schrieb Alain:
> You have "normal" textures.
> You have interior_texture. (an optional element) Add "coincident_texture"
> (as another optional element) with a coincidence_treshold parameter
> (defaulting to epsilon*2 for example) defining the maximum gap that can
> separate the surfaces to be considered as coincident.
>
> For those unfamiliar with the term, the "epsilon" value is the smalest
> value you can have before it's considered as zero.
I know epsilon as the lowest number, where the floating point
representation of the sum of this number and 1 and 1 is greater than 1.
Checking whether to floating point numbers are roughly the same is still
not that easy. If you want to be really correct abs(a-b) < epsilon is not
enough, because epsilon is more like a factor. It all depends on the
calculations done with a and b before.
I think that syntax would get us into a great mess. Is that texture
applied on both sides of these surfaces? Where will the ray be reflected
and transmitted? In my opinion a sort of CSG with more options than
"inside" and "outside" would be better.
Post a reply to this message
|
|
