home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ornl!utkcs2!darwin.sura.net!jvnc.net!rutgers!cmcl2!adm!news
- From: gfo@de-montfort.ac.uk (Gary Fozzard)
- Newsgroups: comp.sys.sgi
- Subject: Re: NURBS basis function on SGI's
- Message-ID: <34006@adm.brl.mil>
- Date: 10 Nov 92 00:47:43 GMT
- Sender: news@adm.brl.mil
- Lines: 83
-
-
-
- On Mon, 9 Nov 1992 C.G.C.vanDijk@nl.tudelft.io wrote:
-
- > > I'm a grad student at UCCS (The University of Colorado at Colorado Springs),
- > > and I'm doing a NURBS project for my geometric modeling course. We're using
- > > some Indigo SGI's, and I'm running into a problem with trying to figure out
- > > what the basis function should be.
- >
- > After three years of working with Nurbs as a PhD student, I now have the
- > opinion that this is an elusive and tricky subject, and certainly not easy.
- > A professional using Nurbs cannot rely on Foley and van Dam alone. Farin's
- > 'Curves and Surfaces for Computer Aided Geometric Design' 2nd (3rd?) edition,
- > Academic Press, is becoming the standard reference work.
- >
- > Anyway, the basis functions of a Nurbs are defined by their (polynomial)
- > degree and a knot partition (i.e. a knot vector and an index where to start
- > on this knot vector). That is it - for everybody, including Foley/vanDam,
- > Farin and SGI. It might be true that everbody uses a different notation, with
- > indices starting at zero or one, or using an 'order' (degree+1) instead of
- > a degree, or implicit or recursive formulas; but all these formulas are the
- > same, and with some maths can be converted to each other.
- >
- > It is common to use a 0/0 = 0 convention, but very careful specification of
- > the start and end value of indices in the recursive formula also works.
- >
- > On your Indigo, you can specify the degree and the knotvector, thereby making
- > any legal B-spline basis function that exists. You can even give weights to
- > your control points, seen by some as relative scaling of the basis functions.
- > The trouble starts when you want to know where to place your control points.
- > That depends all on your application; the field of CAGD has been trying to
- > figure this out for over thirty years now.
- >
- > I do not think that more _real world_ examples in the Iris manuals will solve
- > your problem for you. Look at it this way: the Iris manual tells you how to
- > specify the color for one pixel. But if you do not get into Computer Graphics,
- > you can never write your own ray-tracer by using the Iris manuals alone..
- >
- > Hope I did not put you off :-)
- >
- > Casper
- >
- > Casper G.C. van Dijk "Broken words,
- > Delft University of Technology never meant to be spoken"
- > the Netherlands Bob Dylan
- >
-
- No not at all! I have been looking at NURBS for only one year
- and have found a way to make the perform the way I want about
- 30% of the time! This came about through a lot of hard work
- at the terminal I can tell you. In the end We decided to stick with
- our own suface topologies.
-
- We have an FE model of garment drape which is essentially I triangulated
- grid. We had two options for rendering this which were either direct
- gouruad shading or whatever.. or the interpolation of a NURBS surface
- throught the points in the grid.
-
- I got GOOD results with NURBS for simple shapes but have been only
- 80% there when trying to trim a nurbs surface to look like a garment
- panel! Hence we have dropped NURBS for a bit till I can dedicate
- a year of my life to getting it right.!
-
- Bye the way, in case you were wondering why we ever looked at NURBS
- for this app anyway:
-
- 1) The NURBS library does nice things like autotriangulation of the
- mesh with viewing position.
-
- 2) the trimmed nurbs would allow us to use exact pattern data to
- trim the panel instead of the rather poor resoultion of the FE
- mesh.
-
- Oh well, best get back to it..
-
-
- +--------------------------------------------------------------------+
- + Dr Gary Post: CIMTEX, De Montfort University Leicester, +
- + Fozzard The Gateway, LEICESTER LE1 9BH, UK. +
- + CAD Phone: +44 533 577586 (direct dial) +
- + Cell Email: gfo@uk.ac.dmu (JANET) +
- + Coordinator gfo@dmu.ac.uk (International) +
- +--------------------------------------------------------------------+
-