Path: utzoo!utgpu!news-server.csri.toronto.edu!dgp.toronto.edu!mccool
Newsgroups: comp.graphics.visualization
From: mccool@dgp.toronto.edu (Michael McCool)
Subject: Re: 3-D visualization
Message-ID: <1990Nov7.150717.2240@jarvis.csri.toronto.edu>
References: <1990Nov7.164448.20893@cs.cmu.edu>
Date: 7 Nov 90 20:07:17 GMT
Lines: 64

sfg+@cs.cmu.edu (Sarah Gibson) writes:

>We are looking at vital Image's Voxelview now.  Does anyone know of other
>packages that are available?  Please reply by Email.  

I know you wanted to have me reply by email, but there are some issues 
I wanted to raise here.  I also would like to see a FAQ posting set up;
see the end of this posting.

First, with respect to Voxel View:

MOUNT SOAPBOX

Voxel-View's renderer is a forward-projection volume
rendering algorithm which, as far as I can tell from the images, does
not take resampling into account carefully enough; their images have 
stripes and moire' patterns very often.  To their credit, the images
are produced very rapidly, and can be very useful, but perhaps they 
should look into the algorithm proposed by Lee Westover (farthest 
elaboration in the proceedings of the 90 SIGGRAPH).

The highest-quality images are still produced by a ray-casting algorithm,
since resampling is so much more straightforward.  A naive implementation
of this tends to be slow, but Levoy has shown some acceleration
techniques that are useful (and are not used enough).

Mathematically, of course, they are (almost) theoretically equivalent,
but nobody, including Westover, have been careful enough in maintaining
equivalence.  For example, Westover makes a linearity assumption for
projecting his reconstruction filter which is not totally justified.

DISMOUNT

Anyways.

On a different note, for sites wishing to add their own software or
rendering techniques, something like apE from the Ohio Supercomputing
Center or AVS from Stardent would be the way to go; everything is broken
into modules in a network structure, so you don't even have to "recompile"
to use one of your own modules.  They also encourage a data standard
which is always A Good Thing.

Most of the original messages in this group seem to be "where can I get 
a package to do X".  This could get tedious, so I suggest we set up a 
"Frequently Asked Questions" posting like that used in comp.graphics to 
handle queries of this nature.  I would like to see more discussion of 
algorithms, with possible criticism of systems being produced, and this 
should free up bandwidth for that.

As a start, I think the FAQ posting should include:
	email and addresses for venders of all major visualization packages
		(apE, AVS, Voxel View, ...)
	hardware venders that sell accelerators for volume visualization
	venders for specific markets (e.g. ISG Technologies for 
		medical and life sciences visualization)
	Major references on Algorithms (Levoy's Paper in Trans on Graphics,
	Westover's Paper, etc.), and summaries of the pros and cons of these
		algorithms, along with a truncated description.

I can start to collect this information if no-one else volunteers.  If
someone else wants to do this (or if I am stepping on a moderator's toes
somewhere... I just got here) then let me know.

Michael McCool@dgp.toronto.edu / Dynamic Graphics Project / U of Toronto