View Full Version : MetaNURBS: NURBS or not?

07-09-2008, 07:34 PM
Are lightwave's subdivision surfaces (not catmull-clark) NURBS, or are they a proprietary subdivision surface? I seem to remember them being called MetaNURBS before. I haven't been able to find any information on the underlying mathematical model for these surfaces.

Some features of them seem to suggest the underlying structure may be NURBS, (the fact that they must be quads), others suggest something else (the fact that there are only 4 visible control points per patch instead of 16 for cubic patches.)

I may just be confused about terminology, but I need to find out what I'm dealing with. I'm writing a scene exporter for RenderMan and I need to know if I can just output the control vertices provided through LWMeshInfo as a subdivison surface or if I need to do some translation to make that work right (i.e. translate them into a 16 point NURBS control cage).

07-10-2008, 12:10 AM
Theyre SubDs

07-10-2008, 11:09 AM
If I am remembering correctly, NURBS are like splines, defined by a parametric equation. They are then divided up into quads for rendering.

There is a way to take the verticies and use them as control points for the spline.

LW sub-ds are not too proprietary. They follow basic sub-d rules, AFAIK.

07-10-2008, 11:16 AM
NURBS are splines and stands for Non Uniform Rational B-Spline. :)

07-10-2008, 02:32 PM
First off, thanks for your replies. That clears up my confusion. I haven't used LightWave for a while so I didn't realize that the TAB key actually does something different to the surface than it did in previous versions.

Not that I'm saying anything new to anyone here, but in previous versions of LW your surface had to use quads to use MetaNURBS - apparently LW doesn't use MetaNURBS anymore? So I suppose my question about the old MetaNURBS is still out there.

Anyway, after doing some tests I've discovered that LW's version of SubD's uses a different recursive refinement scheme than Catmull-Clark (which I suppose is an obvious point, otherwise, why two different schemes.) The larger point being that the Catmull-Clark algorithm always results in quads while LW's algorithm seems to result in sub-poly's with the same number of vertices as the original poly.

Why is this important? Well, because the surfaces end up looking very different depending on the control mesh used. The renderer's I'm exporting for only support Catmull-Clark, but if I knew the recursive refinement scheme used by LW I could output raw polygonal data at a given refinement state for LW SubD's and at least get a surface that looked the same, even if it wasn't very efficient data-wise.

So, I suppose I could work out the refinement scheme LW is using (it actually looks pretty straightforward at first glance) but...If someone happened to know what that might be it would save me some headache working it out from scratch. :)

07-10-2008, 02:55 PM
That's interesting, I don't know if the previous incarnation of a renderman exporter was very successful.


Do you think it would be a lot of work to export to the metasequoia format too?

This renderer supports only that format, and it shows great potential:



07-10-2008, 04:42 PM
Okay, well apparently my confusion wasn't cleared up. I talked to a LightWaver friend of mine who's been using LW since the Amiga days. And just so I don't confuse anyone I'll clarify what I was confused about.

Metanurbs used to be the term LW used for their version of subdivision surfaces. The subdivision surfaces are still the same, just a name change. Apparently there were some name copyright/trademark type issues involving the name.

LW's version of SubD's only works on quads and tris but is still the same as it always has been. It's a subdivision surface and has nothing to do with meta or NURBS... Which is what was confusing me.

I thought that the name meta (a concept which is an abstraction from another concept) and nurbs simply meant that the LW programmers were using NURBS as the underlying structure while presenting the artist with only 4 abstrated control vertices instead of the 16 more confusing control vertices required to define a bi-cubic patch like a NURBS.

Shows where thinking gets you right?


That does look promising. I'm not sure how much work it would be to export LW scenes in the metasequoia format. It would depend on a lot of factors. Some formats (3DSMax for example) are hopelessly convoluted. It would also depend on what types of things metasequoia supported and if they could be translated from what LightWave uses (like geometry, camera and render settings etc).

I'm pretty familiar with RenderMan's RIB format, so I'm comfortable writing an exporter for it (not to mention it being dirt simple and very intuitive).

07-10-2008, 06:39 PM
The whole 'meta' thing was part of the old idea of hyping things rather then using industry standard naming conventions. It also applied to 'hyper' voxels as well.