From: broehl@sunee.waterloo.edu (Bernie Roehl)
Subject: Re: SCI: Voxels vs. Polygons
Date: Thu, 6 Aug 1992 13:23:02 GMT
Message-ID: <BsKD6F.4JA@watserv1.uwaterloo.ca>
Organization: University of Waterloo



In article <1992Aug5.054046.958@u.washington.edu> lateiner@cs.unc.edu
(Joshua Lateiner) writes:

>There are many file formats already in existence for voxel data [...]

I guess standard-proliferation isn't a problem faced only by us poly
people!  ("The nice thing about standards is that there are so many to
choose from!")

>[many valid observations about the virtues of voxels deleted]
> Polygons are excellent for making pretty pictures

Which is what current VR is all about.  Most people don't want to use
it for medical imaging, they want to use it to create virtual
workspaces (and virtual playspaces).

>The reason polygonal models are in vogue at the moment is because they
>are (relatively) computationally cheap

Not only computationally cheap, but less bandwidth-intensive than
voxels.  How long would it take to send a complex object using a
voxel-based approach?  Say, a floor lamp (complete with shade)?

>[polygons] are derived from the
>engineering wireframes that have been around almost as long as
>computers have.

And voxels are derived from the pixel-based displays that have been
around almost as long as computers have.  What's your point?

>Only in the past few years has voxel-based scientific
>visualization become a real force in the industry, as computers
>previously did not have anywhere *near* enough power to compute the
>images

I would argue that the vast majority of computers still don't, and
won't for a long long time.  However, I think whatever standards we
evolve should not make any assumptions about the *representation* of
objects; in fact, we should support a wide variety of different
"versions" of each object.  I suspect a CSG model is the best format
for creating objects, since from it you can easily obtain polygon reps
(at a variety of resolutions), wireframes or voxel-based reps.  The
renderer can then ask you for whichever rep it knows how to handle.  A
smart renderer can even ask for the CSG rep (which will be by far the
most compact representation) and convert it at their end to whatever
they want to use.

>Today, with rendering algorithms developed at Lateiner Dataspace Group,
>one can manipulate large voxel worlds at interactive rates on basic
>workstations.

I'm skeptical of this, but willing to try it out.  Where is the code
available?  Either source, or executable for a 486/50 would be fine.

-- 
	Bernie Roehl, University of Waterloo Electrical Engineering Dept
	Mail: broehl@sunee.waterloo.edu OR broehl@sunee.UWaterloo.ca
	BangPath: uunet!watmath!sunee!broehl
	Voice:  (519) 885-1211 x 2607 [work]