[futurebasic] Re:Curve Fits code (was Recursive, etc)

Message: < previous - next > : Reply : Subscribe : Cleanse
Home   : April 2003 : Group Archive : Group : All Groups

From: Robert Covington <artlythere@...>
Date: Mon, 28 Apr 2003 05:41:21 -0400
On Monday, April 28, 2003, at 12:38  AM, Laurent SIEBENMANN wrote:

>
>
> Hi Folks,
>
> rc explained his path approximation problem.
>
>> My needs are...take scanned points and approximate them closely with
>> a curve, reducing the number of points needed to describe the area,...
>
> I'm not sure we have the right leads yet.
>
> Since spline approximation for font creation and
> vectorization has been a major programming activity so
> there has to be a lot of good stuff out there though most
> of it is occulted by copyright and patents.

I have 2 things that I don't mind sending off...maybe a miracle will 
occur.

1 is a near-working example of the FitCurves.c, the "Almost Works" 
project.

Problem is that there (The Almost Works archive), the handle vectors 
are incorrect ,particularly for handle 2...the Error variable must be 
set to huge, and it draws more curves than needed otherwise if not set 
high enough.

The other is the latest and last attempt I have made, a single file 
called MainJuice.bas, which I think is more correct as to all 
parameters, but it quits classic and the Compiler here somewhere in 
GenerateBezier.


I have posted all , including original C sources  to

< http://www.artlythere.com/macsoft/Giving_Me_Curve_Fits.hqx > Stuffit 
will wipe the file type and creator for the Almost Works project file, 
you will need to reset it, it is called simply "Curvy"

Please Ignore comments in globals area that say "Dynamic" when they are 
not and other strangeness that is just commentary Stuff like that. That 
one is my last attempt and has been worked over the eons.

Your duty, anyone's duty, is to peruse, and find the problems. I might 
suggest using the original point set, and then scaling the drawing using

Pointx*40,pointY*40 for instance, in the DrawCurve FN. It seems to work 
better. The original error , gError, for that case is a measely 4 
instead of 6000...6000 because otherwise no curve gets drawn. Set it to 
600 you get 8 curves.

So close, yet so far a cigar away.


At this point, I can't vouch that the general vector FN's are all 
righteous. I  may have edited, copied pasted, fur-balled them into 
wrongness.

At this point, it is a timewarp and a swamp.

Robert Covington