I'd like to suggest that people who make software to run scanners
consider using ICC profiles to maximise the quality of their scans.
This is based on my reading of the MS DDK on ICM v2.0 at
http://msdn.microsoft.com/isapi/msdnlib.idc?theURL=/library/psdk/icm/i
cmstart_5i91.htm
In order to maximise the sublety of colour captured by an input
device, I suggest that the output for each colour captured take up
the
full number of bits available for it, and the range be scalable back
to correct values by an "Image" Device Profile linked into a Device
Link Profile.
Then the Device Link Profile should contain concatenated "Device
Profile" entries for:
- Input device, calibrated by manufacturer, preferably for the
particular machine
- Input media type, e.g. a brand of slide or negative film,
preferably supplied by the film maker
- Input media instance, e.g. a particular film or set of films
sharing common emulsion batch, storage, developing, lighting and/or
other characteristics which affect its colour imaging. This might be
created by a lab doing sensitometric tests, or by using an image with
a standard target created by the camera, or by taking a picture of a
standard target (e.g. Kodak gray scale & colour patches.)
- Image, a profile which will convert the full range stored back to
what was measured.
This should maximise the information in and quality and consistency
of
any captured image, especially when manipulation or output can be
done
with more bits (i.e. more sublety) than the input device can output
linearly.
E.G.
A film scanner can output 8 bits for each of Red, Green and Blue,
i.e.
numbers from 0 to 256. An image has Red [R} values from 23 to 87,
Green from 1 to 222, Blue from 44 to 128. The resulting output file
would have R, G & B values, each scaled (by manipulating the device)
to range from 0 to 256, and an attached "Image" device profile which
would give the corrections to bring the file colours back to image
colours. This assumes that the scanner can measure more unique
values
than it can output, so that the raw range of 64 different values for
R
can actually be output as 256 values, because of the quality of the
A/D converter, or changes to exposure times, or ....
Developers, does this make sense?
Tom
-------------------------
Tom Trottier, Abadax IT Consulting
400 Slater St. Suite 415, Ottawa ON Canada K1R 7S7
__o <TomATrottier@xxxxxxxx> <TomTrottier@xxxxxxxxxxx> ICQ:57647974
_ \< Voice: +1 613 291-1168 fax: 594-5412 No ads or spam, please
(*)/'(*) Vote for your favourite Olympus camera at
http://www.freevote.com/booth/fav_camera
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|