Le Tuesday 28 September 2010 06:49:32 Andrew Ziem, vous avez ?crit :
> On Sun, Sep 19, 2010 at 10:09 PM, stef <stef.dev at free.fr> wrote:
> > Le Wednesday 15 September 2010 15:21:06 Andrew Ziem, vous avez ?crit :
> > ...
> >
> >> Hi Stef,
> >>
> >> Sorry! I had to do two things to make the shading file
> >>
> >> 1. I forgot to define DEEP_DEBUG :(
> >> 2. I ran the commented-out parts of your shell script. (Apparently the
> >> first time I was running the code that was installed with Fedora and
> >> not the code I compiled myself.)
> >>
> >> I am attaching a .zip with the log and all the PNMs. Also there is a
> >> PNG from Windows XP for comparison.
> >>
> >> (The large, solid white area in the Disneyworld PNG I posted on the
> >> mailing list is where I removed a person for privacy reasons.)
> >>
> >>
> >> Andrew
> >
> > Hello,
> >
> > to compute shading correction to be apllied on scans, the backend
> > first scan a white area before the glass. It seems that usable white
> > lines are lesser than other similar models. A short term fix is to try
> > o shorten the numner of lines used. In lexmark_low.c around line 4940
> > change
> > if (yoffset + (8 * 4) / regs[0x7a] < lines)
> > lines = yoffset + (8 * 4) / regs[0x7a];
> >
> > to
> > if (yoffset + (8 * 2) / regs[0x7a] < lines)
> > lines = yoffset + (8 * 2) / regs[0x7a];
> >
> > so we use only 2 lines (you could alos try with 3 lines). If it
> > isn't enough we'll have to change calibration to scan backward to find a
> > better white area.
>
> I'm still seeing the bands with 2 and 3. Should I send the logs again?
>
> Andrew
Hello,
I think you should try what I explained in 'Re: [sane-devel] Colored
vertical stripes with Visioneer Xerox DocuMate 510'. Using another white area
to improve shading calibration seems the best solution.
Regards,
Stef
