Re: searching for an enveloppe

*To*: mathgroup at smc.vnet.net*Subject*: [mg15728] Re: searching for an enveloppe*From*: Daniel Lichtblau <danl>*Date*: Fri, 5 Feb 1999 03:42:23 -0500 (EST)*Organization*: Wolfram Research, Inc.*References*: <78ui46$85k@smc.vnet.net>*Sender*: owner-wri-mathgroup at wolfram.com

Olivier Leroy wrote: > > I'd like to know the envelope of a curve. The data come from image > processing, and are known as a list of points, function of distance. > Does a method already exist to have an upper envelope near by the curve > ? > > Juliette Marais > ___________________ > INRETS > France. If your data points are not misbehaving too badly the following approach might be useful. Assume the curve is approximated by a low degree polynomial of the form p(x,y)=0. (i) Find a least-squares fit to obtain coefficients for p(x,y). (ii) Get an envelope by allowing a small offset, say |p(x,y)|<=epsilon. Daniel Lichtblau Wolfram Research