On Wed, 10 Nov 1999, Automatic digest processor wrote:
> Date: Tue, 9 Nov 1999 10:33:39 +0000
> From: Matthew Pocock <[EMAIL PROTECTED]>
> Subject: Re: Best representation for deformable human models.
>
> You can use a transform to stretch the model. For example, use non-uniform
> scales to make the model wider and shorter without making it thicker. This will
> affect the entire geometry of the person. If you want to be able to enlarge
> just part of it, then break the model up into individualy resizable chunks, and
> scale each of these. The entire body can then be grouped under a single
> transform for global scaling and positioning. Alternatively, capture a
> representative of each body form you want (pear to stick) and use morphing to
> combine the best bits of each.
For nurbs-like deformations, you'll probably want your model, when all is
said and done, to have a regular, grid-like placement of vertices over its
surface. If you over-optimize your model (polygon-reduction, etc.), it
won't reshape well. Aside from morphing and scaling, there aren't a lot
of tools available yet for contorting your figure, so you'll probably have
to figure out the math for some of it, on your own (I'd love to get
together with some folks at some point, and write libraries for bone, IK,
and frame contortions in J3D, but haven't had the time to get it
together). Unfortunately, doing the math to move specific groups of
vertices in your model for deformations is going to be computationally
expensive. Just *how* expensive this is, though, remains to be seen.
Algorithms for deformations can be found in the SIGGRAPH archives. If
those aren't available to you, there are a few other places to look. For
example, a nice OpenGL-based algorithm for simple, smooth joint
deformation can be found in a copy of "Game Developer's Magazine" from
about two years ago. Some nice real-time IK algorithms can be found in
the proceedings of this past year's Game Developer's Conference.
If there is a lot of interest in deforming and/or animating models, I may
be compelled to start an additional mailing list to discuss ideas, and
potentially put together a reusable API or two. If folks are interested,
drop me a line ([EMAIL PROTECTED]). If there's at least five of you, I'll
go ahead and start the list.
----------------------------------------------------------------------------
Tess Snider [EMAIL PROTECTED]
Software Engineer http://www.radix.net/~malkin
----------------------------------------------------------------------------
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
