<%@ include file="../../check-login.jsp" %> Subdivision.org: Re: Real time subdivision surfaces, take II <% session.setAttribute ( "messageNum", new Integer ( 1522 ) ); %>

Re: Real time subdivision surfaces, take II

Posted by Scott Schaefer on February 07, 2005 at 19:13:11:

In Reply to: Real time subdivision surfaces, take II posted by ato3414 on February 07, 2005 at 15:56:17:

: I talked with John Carmack. He has tried (as you mentioned) a Bezier real time subdivision. The consensus was that there is no such thing for subdivision surfaces (this was my original question). All subdivision surface generators take minutes(!) to run. This disqualifies such methods from being usable in real time (i.e. they would need to run in about 10ms). Is there such a thing?

Yes and no. Bezier patches can be viewed as instantiations of subdivision. There exists subdivision schemes for Bezier patches. However, Bezier patches are not extremely complex. A subdivision surface could be subdivided in realtime, but the time taken depends on the complexity of the surface. If you have a lot of polygons or need a lot of polygons as output, it will take a while. If subdivision takes too long for realtime applications for your purposes, you could certainly use subdivision to create a multires version of a model at startup without artist input and use the different cached versions in your application.

There also exsits fast exact evaluation algorithms for subdivision surfaces. Bezier patches are rational polynomials and lead to fast evaluation algorithms. Subdivision surfaces are polynomial almost everywhere. Jos Stam has a paper on a fast evaluation for Catmull-Clark surfaces. The algorithm is a little complicated though and requires storage of pre-computed eigenvectors.

Follow Ups:


Post a Followup

Name:
E-Mail:

Subject:

Comments:

Optional Link URL:
Link Title:
Optional Image URL: