Quad-triangle Subdivision

General Forums Blender and CG Discussions
1

Doing a bit of reading about subdivision surfaces, I stumbled accross this paper about Quad-triangle Subdivision: http://research.microsoft.com/~cloop/qtEG.pdf

It is a hybrid of quad-friendly Catmull-Clark scheme and tri-friendly Loop scheme, and from the sounds of it, it may solve or at least alleviate some of the problems of mixing tris in with a quad-based SDS, such as Catmull-Clark.

I’m no little about SDSs, and much (okay, all) of the math in the paper is above my head, so I’m just wondering if anybody knows anything about this scheme? And if so, does it helps with the problem of triangle distortion in a quad subdivision scheme?

Also, are there any other modellers that implement this scheme?

2

it doesn’t look like it will fix the distortion problem, rather it solves interpolation problems blender’s renderer doesn’t have.

[at least, that is what my quick glance suggested… but the pictures at the end implies this may not be entirely true]

the distortion issue isn’t limited to triangles, for example, picture a 2 by two grid

_______
|  |  |
+--+--+
|  |  |
-------

now imagine that we split apart the second vert and put a quad there… stretched apart we get something like this

_______
|  |  |
|  +  |
| / \ |
+     +
| \ / |
|  +  |
|  |  |
-------

now, if the two verts in the center were brought together by the artist, subsurf would pull their resulting positions apart and change the angles at those corners of the faces… messing up the texturing and bunches of things.

this is what edge loops are for

3

For a closed mesh to be comprised of all quads and only ‘true’ edge loops (all vertices having a valance of exactly 4) it has to be topologically equivalent to a donut. Obviously this is not desirable in all cases…

Cheers
Xarf