gors@well.UUCP (Gordon Stewart) (06/22/89)
I proposed this to Chris Finn, but he didn't seem to like it much.
Consider the following (simple-minded) solution, which does not blow
up in the case of a "bad" set of points, can't generate any floating
point errors, and generally provides "good" results. The goodness can
be quantified after the fact, and for the application I envision --
fitting a circle to a set of hand-drawn points -- is plenty good enuf:
Choose the barycenter of the set of points as the center of the circle;
Choose the mean distance from the barycenter to be the radius.
--
Michael Sierchio (en guise de Gordon Stewart)
{pacbell, hoptoad, pyramid, hplabs}!well!gors gors@well.UUCP