>I am generating points approximating a shape. I want to convert them then
>to curves for more flexibility and smoothness, and ease of storage. Saving
>the knot points for a few curves is a lot easier than listing hundreds of
>points.

So, instead of an equal-distant set of points (i.e., sampling) or the 
actual set of points, you want points that are gathered at the apex 
of each curve. By doing such, you can reduce the number of points to 
approximate the curve. Is that it?

Okay, is the resultant curve recursive (comes back on it's self -- x 
& y = anything); or is it linear (i.e., signal -- y = anything, but x 
always increasing)?

If it comes back on it's self, then finding the apex of each curve 
becomes more complicated because you also have to determine the 
inflection points as well -- like a moving average to find the local 
"maximum" point.

If the curve is linear, then it depends upon what you want to do with 
the resultant data. If you want to just show it -- no problem just 
grab the maximum/minimum y value @ x and your approach is valid. 
However, if you want to later analyze the data, you will find that 
you have destroyed the original frequency and phase information.

So, what type of data and what do you want to do with it?

tedd


-- 
http://sperling.com/