> oliver.rauch@Wolfsburg.DE said:
> > We could create a 3x3 matrix of gamma tables and we would get almost
> > all possible corrections. But the problem is to get the values for the
> > matrix/table for each device.
>
> ICC profiles are (unfortunately) not typically gamma tables. They tend
> to be on the order of 300K files.
If I see the size (300KB) I think that ICC uses a transformation like I descibed
above:
a matrix of gamma tables.
> I think that applying ICC correction
> is out of the realm for the scanner driver, although it may choose to
> arrange for profiles to be associated with the the image.
>
> Hardware manufacturers are typically responsible for producing a profile
> for their hardware, though that is not required, not all do, some get third
> parties to do it, and there is nothing preventing the random individual
> making a profile of his/her own. That implies that the driver need not
> even *carry* the profile.
>
> Typically, the display tool applies the profiles for the scanner and
> the display device to the image at the time the image is rendered. The
> user may even select the profiles to be applied, depending on the tool.
> So this is, and should remain, entirely a front-end issue.
>
So that is what I said, we will not get the profiles for the most scanners.
If we include a color correction into the sane layer as midend, it should be
a correction that is usable for most scanners and does not need big
profiles. So a 3x3 matrix transformation would be the right thing
for the sane layer.
Anything that uses much more data could be included into a frontend
or could be done after scanning.
Bye
Oliver
-- EMAIL: Oliver.Rauch@Wolfsburg.DE WWW: http://www.wolfsburg.de/~rauch
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com