stupid question, subdiv and meshsmooth?

[H3rbie]

Nah, same thing different names.

[beaker]

a subdiv surface is a different type of surface from a polygon. Just like nurbs, polys, etc... are alll their own surface types. Subdiv is a mathmatically smooth surface and just like nurbs it tesselates to polygons at render time(resolution independent).

Mesh smooth is simply smoothing a polygon and the end result is still a polygon surface. If you zoom in and render the edge of the surface, you will still see facets along the edge.

[actarusprocyon]

I thought it would be something along those lines.

So a 'real' subdivision surface would be converting the base mesh's edges to curve similar to a Nurbs' while meshmooth,hence the name,would just quadruples(sp?) the number of faces and not get that 'infinite smoothness' Nurbs have?

If so,why do most major softwares(for instance 3DSm) claim to have subD's surfaces?

Sorry if it makes no sense but I'm new to this.

BTW:I'm learning in XSI Exp if it helps.

[beaker]

>>If so,why do most major softwares(for instance 3DSm) claim to have subD's surfaces?

Max does, it has NURMS(Meshsmooth) and a separate HSDS(Hierchial Subdivision surfaces). Though most people just use NURMS(or whatever they call it these days) because it is faster.

[actarusprocyon]

Ok,probably was I mislead by this thread (http://www.scifi-meshes.com/forums/showthread.php?threadid=320&perpage=15&pagenumber=1)

[beaker]

Originally posted by actarusprocyon
Ok,probably was I mislead by this thread (http://www.scifi-meshes.com/forums/showthread.php?threadid=320&perpage=15&pagenumber=1)
I could be wrong, but I am pretty sure that the 3dsmax's HSDS are true Subd's. The trouble is they are supposedly very buggy, slow and are missing many important features that other subd modelers do have(creasing, direct texturing, etc...). Many of my 3dsmax user friends consider them useless(kind of like their nurbs implementation, half assed). I am guessing that Zaondude is refering to NURMS when he says that they are not true subdivs.

[ThirdEye]

Originally posted by beaker
I could be wrong, but I am pretty sure that the 3dsmax's HSDS are true Subd's. The trouble is they are supposedly very buggy, slow and are missing many important features that other subd modelers do have(creasing, direct texturing, etc...). Many of my 3dsmax user friends consider them useless(kind of like their nurbs implementation, half assed). I am guessing that Zaondude is refering to NURMS when he says that they are not true subdivs.


All the catmull-clark based subds are real ones. There's no such thing as "true" or "false" subd, just different ways of implementing them.

[Dave Black]

I don't mean to butt in here, but I'd just like to put in my $.02

The true definition of Sub-divisional surfaces, may remain elusive:

I have the original math that Ed Catmull & Jim Clark revised based on Doo and Abin's subdivisional rules expressed for the cubic B-Spline surface(which worked on arbitrary rectangular meshes), allowing for extension into meshes of an arbitrary topology.

These methods were then translated into a common system, now know as "Catmull-Clark".

The true term for "Sub-D", really is "Bicubic Recursive Subdivisional Surfaces".

Or,

"Bicubic Uniform B-Spline Surface Refinement".

Newer terms, such as HSDS, all of LW's "special" modeling, I.e., spline surface refinement on a sub-divided mesh, all really translate to the INDUSTRY term, SUB-D, or Sub-Divisional.

In MAX, the MS modifier using the "NURMS" option, is basically, the "Catmull-Clark" method. There is nothing lacking in it's implementation, but there are things ADDED to LW to allow for better manipulation.

At least, this is my understanding, from all my current research on the matter. If I am wrong, I'd love to know.

-3DZ

:D

[beaker]

Here are a few posts from Larry Gritz from his time at Pixar. He consistantly says that subdivision surfaces are not smoothed polygons.


http://groups.google.com/groups?q=gritz+subdivision+group:comp.graphics.rendering.renderman+group:comp.graphics.rendering.renderman&hl=en&lr=&ie=UTF-8&oe=UTF-8&group=comp.graphics.rendering.renderman&selm=7et41k%248ea%241%40sherman.pixar.com&rnum=2
Quoe:
"Stop thinking of subdivs as polyhedra -- they are not. The mesh is a control mesh, just like for B-splines, but without the restriction of being strictly rectangular in topology. The subdivision surface itself is a smooth surface which just happens to be be described by that coarse control mesh.

-- lg"

http://groups.google.com/groups?q=author:lg%40pixar.com+polygons&hl=en&lr=&ie=UTF-8&oe=UTF-8&selm=86sfmb%24fpf%241%40sherman.pixar.com&rnum=10
Quote:
"As an aside, I think we're having a problem with nomenclature here, perhaps because you are working with a package that is polygon-based. Subdiv surfaces are *NOT* made of polygons. They are true curved surfaces, just like NURBS. They are described by a control mesh, just like NURBS, and if you were to draw polygons connecting the vertices of the control mesh, you could visualize that mesh as a faceted surface. But that's the control hull, *not* the subdivision surface itself! Think of a Subdiv like a NURB or a bicubic mesh, except that unlike those types, subdivs aren't restricted to a strictly rectangular connectivity. The subdivision surface is the limit surface you get if you recursively apply a particular set of rules to the mesh. You could visualize the process as one of mesh refinement, but in fact the surface is independent of that (and in fact, large sections of Catmull-Clark surfaces can be directly converted to B-splines).

-- lg"

[Dave Black]

Ok, Beaker, so basically, where does that leave us?

It's all sorta like patch models? Hehe. JK.

Perhaps LW(and I'm simply refering to my very limited understanding of how it's modeler works)does have a "true" sub-d implementation then. Max is pretty much just subdividing polys and averaging their smoothing groups.

Thanks, Beaker. I really feel as though that clears it up.

The sub-d surface is simply the curved surfaced outcome of a hull. It's not polys...at least until rendertime.

I guess that's where all this get's it's meaning. Even though in Max(for instance) you are using a similar technique, I.e, fashoning a low-poly control cage to control a high-poly smoothed(looking) surface.

So Max's "NURMS" method is kinda like a poor-man's sub-d modeling system. For all intents and purposes, it's the same as actually using real Sub-D, but lack the truely res-independant nature of real Sub-d surfaces because it's missing the b-spline ruleset.

From what i can see, at least from your last posted quotes, true sub-d is kinda like the marring of 2 techniques. Like modeling with a low-poly hull, to get a patch or nurbs style surface that is perfectly upwardly scalable because the information is being derived from a spline, instead of just a set of polys. Rather like the difference between raster and vector graphics.

VERY interesting stuff. I can't thank you enough for that one, Beaker. Really.

So am I right now? At least in your opinion?

-3DZ

:D

[FabioMSilva]

is it just me or u said "SUBDIVIDING"? Subdiving is adding more subdivisions to a surface for example adding more smooth level to your meshsmooth, or cutting new lines. But u probably meant subdivision so my post is useless...lol

[Dave Black]

Well, what would you say is the verb for utilizing/activating subdivisional algorithms?

-3DZ

:D

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.