Re: [CinePaint] curves problem in cinepaint-0.22-1

"Peter Langfelder" <[email protected]> Sun, 27 Apr 2008 14:03:04 -0700
Newsgroups gmane.comp.video.cinepaint.user
Message-ID <[email protected]>
On Sun, Apr 27, 2008 at 1:50 AM, Frank Peters <[email protected]> wrote:
> On Sun, 27 Apr 2008 12:03:26 -0700
>  "Peter Langfelder" <[email protected]> wrote:

>  Using my cinepaint on your tmp.jpg original I cannot reproduce your
>  results no matter what kind of extreme curve I apply to the green
>  or value channel.
>
>  Could you possibly post the curve that you have used?
>
>  >From where did you acquire your cinepaint version?
>

Hi,

the curve was very simple - I dragged the top point in the green
channel down to maybe 25%. So the end result was a straight (linear)
curve starting at (0,0) and ending at (255, 64) or so. I jdownloaded
cinepaint 0.22-1.tar.gz this morning (04/27/2008) from sourceforge.
Had to fix a call to something in fltk to get it to compile (the fix
was discussed on the mailing list recently).

In the meantime I dug into curves.c and found the problem. In function

static void
curves_plot_curve_u16(

which, as far as I can tell, calculates the curve between given
points, there is a constant 70000 steps that are calculated. I changed
this to 200000 and voila, the problem disappeared. I'm not an expert
but my feeling is that the steps (dx, dx2, dx3, dy, dy2, dy3) needed
to be finer for the calculation to work properly.

The problem may not appear if you define several points on the curve,
because the 70000 steps then apply to each segment and that may be
enough.

Thanks and best regards,

Peter

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone