On Mon, 26 Jun 2000 22:05:15 +0100, "Mick Hazelgrove"
<mic### [at] mhazelgrovefsnetcouk> wrote:

>Care to explain in english!!!!

About the tileable pigment, Chris Colefax has a macro that would make
any texture tile seemlessly. Granted, a repetition *may* be noticed
but the probability would not be as high.

Now about the bounding quadtree. Imagine you have a big (and
relatively flat) grass patch consisting of, say, 32x32 smaller patches
whose bounds you know. Bound each patch into its own bounding box. You
now have 32x32 bounding boxes. Bow bound them in quads, 2x2 so that
you have 8x8 bounding boxes each of which has 2x2 bounded patches in
it. Then do this again so that you have 4x4 larger bounds, then again
to get 2x2 and finally bound them into a single box.

Why would you want to do that?

Imagine a ray-shape intersection test. A ray is shot from the camera.
If no bounding is used, it would have to be tested against each of the
32x32=1024 patches (consider them single objects for simplicity). With
a hierarchical bounding scheme such as the one described above, the
ray will first be tested against the outermost box, then if there's an
intersection it will be tested against the inner four and so on and
the test will traverse down the quadtree (quad because each 'branch'
has four sub-branches) and only follow those branches which are likely
to result in an actual intersection.

The best results would be achieved if the grass field is viewed from
above because the projections of the bounding boxes on the screen (the
vista buffers) would not overlap. I am sure a more efficient method
can be developed which would incorporate adaptive median subdivision
of the bounding boxes combined with minimizing the vista buffer
overlapping but I would probably leave this for povray.programming or
povray.unofficial.patches ;)

Hope it helps any.


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] usanet
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message