Don't forget about the peculiarities of the human eye too. Even when
scanning with monochromatic light, what Nikon Coolscanner do with
their RGB LEDs, the colors sometimes don't look 'right' at all.
The main issue when doing photographic scans is not to capture
the EM spectrum as accurately as possible, it's to capture
as well as possible the scene as it appears to a human.
> A red signal is generated when the frequencies that the sensor respond
> to, the frequencies that the illuminator emit, and the frequencies that
> subject reflect (or transmit, for transparencies) all match. It gets even
> more complicated in some scanners, that have color filters in the optical
> path. These are sometimes used to compensate for issues wrt illumination
> or CCD sensitivity.
>
> The closest thing there is to a calibration table for scanners is
> an ICC color profile.
Luckily scanners are linear devices so we don't need LU tables; A 3x3
color transform matrix is all that is needed. This color transform
happens in the scanner right now but I would like to have the matrix
exported as a pseudo well known SANE option as well. It would have to
transfrom from scanner RGB to a well defined linear colorspace,
preferably to the space defined in Rec. 709 for HDTV, only not with a
gamma of 2.2 but with 1.0 (i.e. only using the Rec. 709 primaries and
white point).
-- -- Ewald
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com