Breton Slivka wrote:

> This is a vague idea, from a vague understanding, and you're probably 
> already way ahead of me on this...
> 
> as I understand it, it is a fairly simple matter to subdivide a 4 point 
> patch into two 4 point patches, by just substituting boundary points, 
> with control points, and interpolating intermediate control points. 
> (Which in itself would be a nice feature in a patch model, to arbitrarily 
> subdivide a patch)
> 
> Anyways, with the paper on 3 5 and 6 sided patches, you are provided with 
> mathematical definitions (A bit hard to swallow) for 3 and 5 points 
> patches, along with patterns for control points. 
> 
> Couldn't the same principle for subdividing 4 point patches into 2 4 
> point patches be applied here? You could start with a 5 point patch 
> definition, and then subdivide into  5 4-point patches by substituting 
> control points in the 5 point patch definition into boundary edges for 
> new 4 point patches, and interpolate the individual control points in the 
> same manner you would when subdividing a 4 point patch?
> 
> Stop me if this all sounds stupid, or obvious...

Let's use the five-point patch as an exaple. I'd like to subdivid it in 
a way where a center (of the patch) connects to the center of each 
edge-curve-segment - this way I'll get 5 4-sided patches (see the paper 
by Martin Hash I mentioned in my previous porting for a diagram).

Subdividing the edge-bezier-curves is a simple de Casteljau operation - 
I've had some problems with finding the patch center, put the paper 
posted by Rune has the solution for that (at least I think so). The only 
problem that needs to be solved is finding the control-points for the 5 
new curves that origin from the center. And - all in all - I'm not sure 
that those five new patches will connect 100 percent smoothly.

All I can do is playing around with patches by trail and error, but it 
would be better to have a sort of a mathematical proove, but that's 
beyond my means...

-sascha

Post a reply to this message