On Tuesday 25 May 2004 02:28, Kai-Uwe Behrmann wrote:
> Am 24.05.04, 10:58 -0700 schrieb Troy Wu:
> 
> > Unfortunately, being able to measure output values only solves half 
the
> > problem.  For a full color-correction workflow, I need to be able 
to
> > soft-proof, which requires that my monitor is profiled and 
calibrated.
> > To do so requires a colorimeter--or spectrophotometer--which can be 
used
> > by Linux.
> 
> It seems to be the only possiblity at the moment to use the wine 
emulator,  for running calibration applications.

This is true but you also need every part of the work flow to be color 
aware or at the very least have software that implements the various 
manipulations needed.  

For example in Windows I am using the SilverFast software for my film 
scanner front end.  In SilverFast there is a color management dialog 
in Preferences that allows me to specify my CM workflow for scanner 
operations.  So I can specify my scanner device profile, my working 
space profile and other information about my workflow.  Then when I 
scan an image the result is a file that has been scanned, then 
characterized with the scanner profile and then tranformed to my 
working space profile.  What I get in PhotoShop is a correctly 
characterized scan already converted to my working color space.   So 
once this is setup I don't even have to think about it.

LCMS will let me do the characterization and transformation steps but 
it does require additional steps on my part for every image I scan.  
If your scanner is supported by VueScan, one of my is and one is not,  
then you can have similar facilities in Linux.  But as nice as VueScan 
is it is not open source. 

In Windows the CM workflow is fragmented and in order to implement a 
working CM workflow the user must enter configuration information in a 
number of different places.  For each input device front end,  in the 
image editing software and in the printer driver for each 
paper/ink/resolution.  This fragmentation makes this a difficult task 
for many users mostly because it is difficult to pull up configuration 
information for the workflow in one location and it is therefore 
difficult for the user to visualize the overall process.  For the user 
with color management experience who understands how this works and 
where all of the pieces are it is manageable.  But the learning curve 
is significant and learning about color management and how to 
configure it at the same time has a very steep learning curve.

This ignores the additional  problems involved in characterization of 
the devices used in the workflow - cameras, scanners, displays and 
printers.  At least in Windows the software and hardware to do the 
characterization of these devices exists.  But even with these 
characterization facilities this can be an expensive and difficult 
task.  Display and print systems are particularly difficult to 
characterize.  Characterization of scanners and cameras is a 
relatively simple task in Windows and should be in Linux as well.  

> It seems to be the only possiblity at the moment to use the wine 
emulator,  for running calibration applications.

One other point about the characterization process.  This process 
REQUIRES that every part of the chain involved be EXACTLY the same 
when doing the characterization as it will be in actual use.  For 
example when profiling a printer you must so everything exactly the 
same when printing your target as you will during your normal workflow 
other than selecting a printer profile ie there is no transformation 
between color spaces.  So everything matters; driver settings 
(resolution, paper type, dithering...), paper, ink - everything.   So 
using wine will work for the characterization of scanners, cameras and 
printers since the normal Linux workflow can be used to capture/print 
the target and the Windows software is only being used to generate the 
profile from the captured data.  But LCMS will handle this so there is 
no need for wine to profile cameras, scanners and printers (maybe not 
printers).   But I have serious reservations about doing this with 
displays because there will be an added Windows software layer in the 
display chain during the capture of the raw characterization data that 
could, very likely does, alter how colors are displayed.  This will 
result in the generated profile being incorrect or more to the point 
only correct for software running under wine.   

The GIMP only needs slight changes to make it a usable, not ideal but 
usable,  part of a CM workflow.  It currently has a color proof 
display filter that allows the user to select a ICC profile for soft 
proofing.  I belive that they have used LCMS to implement this 
feature.  All it needs is the same feature for a display color space 
profile, the ability to use the images embedded profile and the 
ability to apply a user selected profile to an image on its way to a 
printer.   These changes would make The GIMP usable in the short term 
until a more comprehensive open source system wide CM system is in 
place.
 
I like Bob Friedenhahn's idea of having one place that contains all of 
the CM configuration data.  However,  I think the correct 
generalization would be to allow sections for multiple display 
devices, multiple input devices and multiple printers/output devices.  
After all how many of you only have one scanner/camera or one printer.  
Bob's spec handles multiple display devices.  The names of the 
sections would be the same as those used by the software that controls 
the device.  So the display section names would be the same as the 
display names in FX86config.  The config software for each device 
would read/update this file in it's color management dialog and the 
driver/editing software would read this file to implement it's color 
management processes.   

In addition this config file could/should contain CM workflow related 
information not just characterization information.  This is a 
significant and logical extension of Bob's idea.  But it is also 
non-trivial and might better be left as a "phase 2" item.  Since 
having all the system CM characterization configuration data in one 
place is a huge step forward and is a significant improvement over 
Windows.  Anyone here worked with CM on the Mac?  How is this handled 
there?  Perhaps there are best practices and things to avoid that can 
be learned from the Mac that should be applied to the open source 
world.

As to the performance of LCMS.  I personally am not concerned that it 
might be a little slow.  What matters at this point is that it has a 
well designed api and that it works.  Performance issues can be 
corrected over time.  On my list of things that need to be done for CM 
on open source platforms this is near the bottom of the list.  First 
get things working then make them perform better.  LCMS works we need 
to get the other parts of CM working.  

-- 
Hal V. Engel

Attachment: pgpTUmjX2K1CQ.pgp
Description: signature

Reply via email to