Splines VS Polys 2005-03-07 - By nicholas8681
Back
Now I'm not talking about the spline modeler, I'm talking about TRUE
spline modeling. Ala. spatch, hamapatch, jpatch, animation master, etc.
Is one more efficient then the other. An interesting little articles
from Martin of Hash, INC.:
"Hash patches have the tremendous advantage that they do not need to
be aware of the patches that surround them. Each patch is a
self-defined, high-order surface that can be manipulated without
knowledge of its neighbors. The normal anywhere on a patch's surface
can be determined analytically. This is very important for some of
Animation:Master's character animation features, especially Muscle
motion and Smartskin, but it is also important for any movement that
distorts the spline mesh.
In contrast, polygon models are in reality made of many flat surfaces
that only appear smooth because the normals where polygons abut are
averaged together. Information must be maintained of where polygons
join, called "vertex" and "edge" lists, and whether and how much
normal averaging to do at each vertex. List maintenance requires a
significant amount of preprocessing. Without normal averaging, every
polygon would "crease" with every other polygon, however, people often
positively compare a polygon model's simulated "smoothness" to a
patch's true surface because a patch's high-order surface may not
appear as "smooth" as the simple linear interpolation used by polygons.
The "porcelain.mat" material is a solution to achieve polygon-like
"smoothness". One technique would be to divide the patch into a grid
of polygons, then use traditional polygonal vertex and edge lists to
determine averaged normals, but this would require burdening overhead.
Instead, the cubic nature of the Hash patch technology offers a
better, and more accurate alternative. Given four normals, a Hermite
cubic can determine any normal along a line. Four Hermite cubics,
combined via Coons interpolation, can determine a normal anywhere
inside a rectangle. These are not the true normals of the underlying
patch, and they often fail to adequately represent high-order shapes,
but they do appear "smooth" in a polygon sense.
The initial 12-normals are indeed the "real" normals of the patch, so
in theory, the patch is still independently definable. However, if two
patches were modeled with a "crease" where they joined, it would still
be there. Therefore, when porcelain.mat is applied, a computationally
expensive preprocessing step is performed to determine possible
creases by examining a patch's connectivity with other patches, and
adjusting the normals accordingly. This masks patch modeling creases
much like the creases where polygons join is masked.
Another option is also available when using porcelain.mat. Since a
patch using porcelain knows its neighbors, it can blur normals with
them depending on the "weight" setting of the porcelain material. A
weight of "0" does no blurring, (however, porcelain normal
interpolation is still used), while a weight of "100" will blur
equally the current normal with the four normals two-thirds of a patch
away, (a weight of "50" will average one-half the average of the four
surrounding normals with the current normal.)"
The article is discussing a new feature that was implemented into
Animation Master at the time. The question is, are splines better for
modeling or are polys? I'm not talking about personal preference? I'm
talking for computational purposes, ease of modeling, and intelligent
modeling. Reading up on polys and splines, it seems that polys ARE a
antique way of modeling because you are essentially creating a smooth
figure by making a large amount of tightly night hard edges so that
the object appears smooth but in reality isn't. While splines, are
truly organic.
If this is the case, why hasn't spline modeling been more used and
implemented, I know it is in many "highend" programs like lightwave,
maya, etc. But it seems to rarely used, is it lack of knowledge of the
features or just old habits die hard?
Just striking up a debate, and curious what others think.
Brian
Welcome Yahoo Members!
http://www.eovia.com
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/Carrara/
<*> To unsubscribe from this group, send an email to:
Carrara-unsubscribe@(protected)
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|
|
