plustek@linuxhacker.org

Hi,

On Sonntag, 4. Januar 2004 14:03, Alain Wenmaekers wrote:
> I have experimented a bit with my Canon Lide 20 and sane 1.0.13.
>
> A few months ago, I had a discussion on this mailinglist about bad color
> quality with this scanner (too much red in the picture).
> After playing with the (manual) gain values, I got a all colors scanning,
> but still in bad quality.
>
> I wanted to try sane 1.0.13 because of the "new calibration" code, to see
> if anything was different.
> Unfortunately, I still have red color scans (the only color that get
> scanned is red).
>
> Then, I changed a piece of code in sane 1.0.13:
> line 648 in plustek-usbdevs.c
> Original: {{ 0, 110}, 35, 10, {2550, 3508}, {75, 75}, COLOR_GRAY16 },
> Changed To: {{ 0, 110}, 25, 10, {2550, 3508}, {75, 75}, COLOR_GRAY16 },
> -> Third value from 35 to 25

Good idea ;-)

> After compiling, I got this scan:
> http://users.skynet.be/bluedigit/xsane.jpg
> reference scan (vuescan): http://users.skynet.be/bluedigit/vuescan.jpg
>
> As you can see, every color is there in the right proportion, the only
> problem is a dark picture. The preview mode of "xsane" displays an
> identical picture as vuescan when setting the "brightness" higher by 60
> points.
> Here's a scan of the brightness set to +40 points:
> http://users.skynet.be/bluedigit/xsanebrightness.jpg
>
> The same problem occurs with gray/b&w scans too. Brightness is too dark by
> the same margin.
>
> What I can understand from the code, the changed value is a reference where
> the calibration strip is located. The only thing that still goes wrong is
> the brightness/contrast calibration.
> Why is this value 35 while my scanner only responds to 25? (I understand I
> may have some exotic version of the Lide 20... but it's still fun to find
> the reason why sane is affected, while vuescan is not).

The 35 is a distance, which works good for me here on my N670, which should
be the same device like the LiDE20. So 35 is more or less a experimental 
value. In the end, there's no brightness/contrast calibration. During the
calibration step, we only can adjust the white and the dark-points of each
channel and of each pixel, the so called coarse (for the analog front-end) and
fine calibration...
When theses points are found, the only way you can influence a scan is
adjusting the "gamma" curve - a function which tells the scanner how each
incomming value is interpreted. The default gamma-curve is a linear function.
Probably some adjustment there should also be done per default...
All this stuff and probably an autodetection of the "white" calibration strip
is done by vuescan automatically - you might ask Ed Hamrick ;-)

Anyway, don't tell me such things like "vuescan is able to do it, the
SANE-backend not...". Ed Hamrick sells his software and he did a good job,
but always remember, that the SANE-backends are written on spare time,
often enough without the help of the manufacturers and they are for free.
Believe me, if I had enough time, I'd like to make it more prefect....

Okay enough now ;-) I'll do some more tests, and hopefully we can fix this
issue ASAP...

Thanx for your informations and tests,
  Gerhard





>
> Any ideas are welcome...
>
>
> Grtz
>
> Alain
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: plustek-unsubscribe@linuxhacker.org
> For additional commands, e-mail: plustek-help@linuxhacker.org