25 Nov
2003
25 Nov
'03
1:22 a.m.
Bob Jamison wrote:
What I would like to do, is for each candidate pair, ignore the center point and consider the path to be a single 4th degree curve. This curve would then be reduced in degree to make the 3rd degree approximation at (B).
I can't help but think that this is a solved problem with a much more elegant solution. I think bezier's are essentially bernstein polynomials, so there is undoubtably lots of work on adjusting approximations. Having said that, I haven't been able to find anything remotely helpful.
Conceptually it seems that you could replace 3 segments with 2 with a smaller error term, or 128 with 127. There may be a MDL type algorithm for it too.
Keep looking though :)
njh