[sane-devel] coolscan color problem

Hubert Figuiere hfiguiere@teaser.fr
Wed, 8 May 2002 10:57:44 +0200

Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

According to Major A <andras@users.sourceforge.net>:
> > -from negative film, color is really weird. This is the positive
> > image, but they really don't match. I selected "Kodak negative" as
> > media type.
> Colour negatives need special treatment in software to eliminate the
> orange mask. I don't know what xsane does in this area, but I guess it
> cannot be perfect simply because the algorithm should ideally be
> applied to the unprocessed data from the CCD rather than the one
> processed through gamma tables etc.

The resulting image (postive) has an orange color color cast. It
should not because the negative of orange is not orange....

> > -from black and white negative, picture is really dark, and really
> > grainy, and no that is not the film grain. I selected "Slide" and told
> > to invert after scanning, because I don't have B&W negative type.
> I think you will find satisfactory results if you play around with the
> exposure and gamma options a bit.

I used auto exposure and automatic gamma.

> > -from slide film, I still get grainy pictures (even though I'm
> > scanning Provia or Velvia slide, both really common and one of the
> > most fine grained films). I selected "Slide" as media type.
> Is this really grain or just noise? In the latter case, try increasing
> exposure, which will improve signal-to-noise ratio.

Noise. I thought that auto-exposure would give me the right settings.

> > In the three cases, the contrast is excessive. This is even more
> > visible on slides taken by real sunny day.
> What gamma settings are you using?


> > Is it a known issue ? If yes, what can I do to help resolve the
> > problems ? I can do whatever you want including debugging, coding and
> > taking pictures :-))
> Have you tried VueScan on the same pictures? If you manage to get good
> scans from it, then it is a software problem.

I can't because I'm on Linux PPC (VueScan only works on Intel Linux)
and I can't plug the scanner to my PowerBook that runs MacOS X before
I get this USB<->SCSI adapter....

> On the other hand, coolscan2 currently allows you to set all relevant
> scanner options, and with 10bit output you can get exactly the same
> information as VueScan would give you. Therefore this all seems to be
> a setting and/or postprocessing issue rather than a problem with the
> backend.
> > I'd really like to use my coolscan with SANE.
> One last question: did you use coolscan or coolscan2?

I used mostly coolscan2.

Thanks for the tips.

Content-Type: application/pgp-signature
Content-Disposition: inline

Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org