[Tfug] Color Management

[Brian Murphy]

Quoting Bexley Hall <bexley401 at yahoo.com>:
> Unfortunately, without an understanding of what is actually
> involved in the technical process, its hard to decide (a priori)
> whether there is any *real* difference between this paper and
> "some stuff I found lying on the floor in the closet"  :-/


You would need to print a test pattern on your "other" paper and use a
colorimeter to profile it to make it approximately the same.

Your question of allowed tolerance between paper A and paper B isn't
something TFUG can answer for you.  (e.g. good enough for gov't work)


>> combination, and in my experience, these are specific to
>> photoshop.
>
> Not sure why that should be.  I thought the profiles were
> application independant (or, is photoshop the only application
> that fully *implements* this?  e.g., does the application
> massage the data as it is being printed/displayed?  or, is
> that offloaded to the associated driver??)

The application needs to be aware that rgb(10,30,102) detected from the
scanner should be displayed as rgb(9,30,105) by the monitor for the
colors to be the same in absolute terms.

Apps that don't implement profiles won't do the transformation.


>> You can buy a calibrator to generate a printer profile, but
>> these tend to be much more expensive (>$1000)
>
> Hence the suggestion of using a calibrated scanner to
> calibrate a printer.  I.e., if I scan a known target,
> I now know how to compensate my scanned images to reflect
> "reality".  So, if I *print* a known *source* and then
> scan *that*, I can figure out how "wrong" my printer is.

A colorimeter is basically a calibrated camera/scanner so that should
would do it if you had software that allowed it.


> Or, is the process so imperfect that the errors add up
> REALLY FAST?

Your scanner would need a wide enough gamut to cover the colors produced
by the printer for best results.


>> > Lastly, can profiles be "convolved"?  E.g., if you scan something
>> > on an unprofiled scanner, can you later convolve P(scanner) with
>> > P(printer) at print time (i.e., redefine P(printer) as such)
>> > to get the same result as if the image had been scanned on the
>> > *profiled* scanner?
>>
>> The end to end process requires calibration all along the,
>> and devices
>> capable of reproduction of the same colorspace as the
>> original.
>
> But, I don't see how that *should* be the case.  I.e., I would
> imagine that each "profile" is just a "transform" of sorts that
> maps an input image into an output image.  So, you should be able
> to create a *composite* profile that goes through N steps.

Mathematically you should be able to create a single transformation map
end to end if you know mappings for every device does along the way and
data is in range.  Once you throw in a device that is unprofiled, you
have no guarantee what the result will be.  For starters, you don't
know the gamut for the unprofiled device.

In your specific case of an unprofiled scanner, you don't know how the
scanner recorded the true color of the source image.(extreme case,
blues are saved as purples)  Thus it doesn't matter that you have a
calibrated monitor and printer.  Garbage in, garbage out.

Brian

The opinions or statements expressed herein are my own and should not be
taken as a position, opinion, or endorsement of the University of
Arizona.