Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!think.com!mintaka!bloom-beacon!eru!hagbard!sunic!mcsun!inesc!eniac.inesc.pt!mag
From: mag@eniac.inesc.pt (Manuel Noronha Gamito)
Newsgroups: comp.graphics
Subject: Help on raytracing proc. defined fractals.
Keywords: raytracing, Perlin's Turbulence.
Message-ID: <1727@inesc.UUCP>
Date: 30 Mar 91 17:06:20 GMT
Sender: news@inesc.UUCP
Lines: 34
Nntp-Posting-Host: eniac.inesc.pt

Hi, guys!
Perhaps somebody can help me about this:

I'm trying to raytrace a procedurally defined height field using
Perlin's turbulence function, wich has fractal characteristics.

The key problem in this is to find a good ray-object intersection
algorithm (I guess this is the most discussed issue about raytracing :-)

A rather obvious solution is to discretize the turbulence function into
a two-dimensional array and use standard methods for tesselated 
height fields like Musgrave's gridtracing.

I thought of an algorithm that doesn't use that much dynamic memory
without being too slow (... I hope):

1- Make the ray go trough a hierarchy of cheesecake extents surrounding
progessively smaller areas of the fractal surface.
The process stops when a leaf extent is found surrounding an area small
enough so that the surface normal is known to be well behaved inside it.
2- Apply Newton's root findind method to calculate the ray-surface
intersection. If the area found in 1 was correctly calculated there
won't be any stationarity points in it's interior and therefore Newton's
method is guaranteed to converge.

Could anybody point out possible weaknesses, give sugestions, direct
me to good papers on this topic, ... you name it.

Thanks!
-- 
Manuel Noronha Gamito, INESC          		| email: mag@eniac.inesc.pt
"There are billions of planets in our galaxy.   | mcsun!inesc!mag@uunet.UU.NET
Why should this modest planet be the only one to have   | Phone: +351 1 545150
living forms? The Cosmos must be filled with life!"     | Fax:   +351 1 525843