Tom Fogal wrote:
> 
> I was curious if POV ray itself or anyone out there was working on an
> AMR based representation of the objects in a scene. For those not
> familiar with the buzzword, its basically the same old subdivision
> methods except now the tree can adapt in time. For example, a ball in
> the upper left quadrant might be flying towards the upper right in an
> animation. At the beginning of the animation, the top left quadrant
> would have lots of resolution -- that quadrant would be defined by 4
> subquadrants, and one of them (the one that contains the ball) might
> have 4 of its own divisions.

You should know this only makes very limited sense since POV-Ray shapes 
are not approximated with triangle meshes for the render.  Your 'ball' 
would be just a sphere which is a single entity and does not profit from 
being subdivided in a spartial data structure.

Also note the required detail level does not only depend on the distance 
of the camera in a raytracer - adaptive subdivision (for thse shapes 
where it could make sense) would only work on per-ray basis.

Finally in contrast to a scanline renderer you don't gain a lot of speed 
by replacing a detailed triangle mesh in a raytracer with a low detail 
representation.

So to sum it up the only significant gain of such a technique would be 
cases of mesh subdivision where it is either very expensive to generate 
the whole mesh or it takes too much memory to store it.

Christoph

-- 
POV-Ray tutorials, include files, Landscape of the week:
http://www.tu-bs.de/~y0013390/ (Last updated 01 Jul. 2005)
MegaPOV with mechanics simulation: http://megapov.inetart.net/

Post a reply to this message