tweak/refine
Summary
Refine creates an abstract
profile, by comparing CIE measurement values from two test charts.
The charts will usually be in .ti3
format, but only XYZ, Lab or spectral values will be used (ie. all
device space values are ignored). Typically the charts would be
printed on a target system (the one being emulated, say a printing
press), and the proofing system (the one that is being profiled).
The abstract profile that refine
produces will be a correction that makes the proofing system behave
more like the target. This can then be used to recreate the proofing
systems ICC profile, or device link. By feeding a previous abstract
correction profile in as well, iterative improvement can be made to
the proofing reproduction.
verify is a useful tool to use on the two
test charts, to check how well the refinement is proceeding. If a
white point relative match is being created (refine -R), then use
veryify -N.
Usage Summary
usage: refine [-options] cietarget ciecurrent [outdevicc]
[inabs] outabs
-v
Verbose
-c
Create
initial abstract correction profile
-g
Don't
impose output device gamut limit
-r res Set
abstract profile clut resolution (default 33)
-d factor Override default damping
factor (default 0.950000)
-R
Aim
for white point relative match rather than absolute
-f
[illum] Use Fluorescent Whitening Agent
compensation [opt. simulated inst. illum.:
M0, M1, M2, A, C, D50 (def.), D50M2, D65, F5, F8, F10 or
file.sp]
-i illum Choose
illuminant for computation of CIE XYZ from spectral data
& FWA:
A, C, D50 (def.), D50M2, D65, F5, F8, F10 or file.sp
-o observ Choose CIE Observer for
spectral data:
1931_2,
1964_10, S&B 1955_2, J&V 1978_2 (def.)
cietarget
Target CIE or spectral values, CGATS file (e.g. .ti3)
ciecurrent
Actual CIE or spectral values, CGATS file (e.g. .ti3)
[outdevicc]
Output device ICC profile to set gamut limit (not used if -g)
[inabs]
Previous
abstract correction ICC profile (not used if -c)
outabs
Created/refined
abstract correction ICC profile
Usage Details
refine provides a way of improving the profile accuracy of a
proofing system.
The -v flag prints out extra information during the
checking, and prints each patch value, rather than just a summary.
The -c option is used when refine is being used for the
first time, and there is no previous abstract profile to continue
refining. If -c is used,
then the name of the previous abstract correction profile should not
be supplied.
If the -g flag indicates that an output device profile is
not being supplied, and that corrections should be attempted, even
if the colors are outside the devices gamut. Normally an output
device profile is supplied, and corrections aren't applied to colors
outside the devices gamut, since this will not achieve anything
useful, and can distort the results.
If the -r parameter overrides the resolution of the CLUT
grid used in the abstract profile. By default the value is 33, but
other
values can be chosen. An odd number is recommended.
If the -d parameter sets how aggressively refine should try
and correct errors. Normally it will try and exactly compensate for
the color errors revealed in comparing the two measurement files,
but if the device behaviour is unusual, or not very repeatable, this
may result in successive applications of refine making things worse,
rather than better. If this is the case, try using a smaller number,
such as 0.8, or 0.5.
If the -R flag is used,
then refine creates an abstract profile for improving the match of
the patch values when interpreted in a white point relative
(ie. Relative Colorimetric) intent. If used to create a corrected
device link profile using collink,
remember to create a Relative colorimetric intent device link
profile.
The -f flag enables Fluorescent Whitening Agent (FWA)
compensation. This only works if spectral data is available and, the
instrument is not UV filtered. FWA compensation adjusts the
spectral samples so that they appear to have been measured using an
illuminant that has a different level of Ultra Violet to the one the
instrument actually used in the measurement. The optional
illumination parameter allows specifying a standard or custom
illumination spectrum to be used as the similated instrument
illuminant, overriding the default D50 or CIE computation
illuminant used for FWA (see -i below). See colprof -f for
a fuller explanation. The same value should be used as was used
during the creation of the profile.
The -i flag allows specifying a standard or custom
illumination spectrum, applied to the spectral test point values to
compute CIE tristimulus values. A, D50, D50M2,
D65, F5, F8, F10 are a selection of
standard illuminant spectrums, with D50 being the default.
If a filename is specified instead, it will be assumed to be an
Argyll specific .sp
spectrum file. If FWA compensation is used during measurement, this
illuminant will be used by default as the simulated instrument
illuminant. The same value should be used as was used during the
creation of the profile.
The -o flag allows specifying a tristimulus observer, and is
used to compute CIE tristimulus values. The following choices are
available:
1931_2 selects the standard CIE 1931 2 degree
observer.
1964_10 selects the standard CIE 1964 10 degree
observer.
1955_2 selects the Stiles and Birch 1955 2 degree
observer
1978_2 selects the Judd and Voss 1978 2 degree
observer
shaw selects the Shaw and Fairchild 1997 2 degree
observer
If both CIE and spectral values are present in the input files, the
CIE values will be used by default. Using the -i, -o or -f
flag will force spectral values to be used. The the -i, -o or -f
flags will apply to both the target and measured input files.
cietarget
Is
the filename of the target CIE or spectral values. This is a CGATS file (e.g. a .ti3 made using chartread). These are the color values
wanted for each patch in the test chart, typically the product of
the target print system.
ciecurrent
Is the filename of the
actual, current measured CIE or spectral values. This is a CGATS file (e.g. a .ti3 made using chartread). The errors between these
patches and the patches in the cietarget file will be used to create a
correction profile.
[outdevicc]
If
the -g flag is not used,
then the output device ICC profile should be supplied here, to allow
refine to limit its
corrections to colors that are within the gamut of the device.
[inabs]
After
the first correction has been created, subsequent corrections need
to improve upon previous ones, so the previous correction profile
should be provided here. For the first correction, the -c flag should be used, and this
argument is omitted.
outabs
The
name of the created or refined abstract correction ICC profile
Discussion
Refine is typically used in
a proofing situation, in which a verification chart is being used to
check the accuracy of a proofing system. (It might also be used for
more arbitrary color alterations by created two test chart files by
hand.) By using the errors between the target chart and the measured
values, refine attempts to improve the match between the proofing
system and its target.
There is facility in collink, colprof and revfix
to incorporate an abstract profile.
For systems using two device profiles or a device link to convert
between the target space printing files and the proofing device
space, the following would be a typical scenario:
We have a reference set of test
chart values, read from the target system reference.ti3. The ICC profile
for the target system is target.icm.
The ICC profile for the proofing system is proofer.icm. If using a device
link, the device link used to print proofer test charts is
currently target_proofer.icm:
First we print the test chart out on the proofing system and read
it in, resulting in a chart1.ti3
file.
Lets check how well the proofing system current matches using
verify:
verify reference.ti3 chart1.ti3
We then create our initial abstract correction fix profile fix1.icm using refine:
refine -v -c reference.ti3 chart1.ti3
proofer.icm fix1.icm
Applying this to your process for creating the proofer device
profile or target to proofing device link (choose one of the three
options below, depending whether you are using the proofer profile
and just want to alter its colorimetric B2A table using revfix, whether you are going
to recreate the proofer file from the original measurement data
using colprof, or
whether you are using a device link profile created using collink):
revfix -v -1 -ke -p fix1.icm proofer.icm
proofer_fix1.icm
or
copy proofer.ti3 proofer_fix1.ti3
colprof -v -p fix1.icm proofer_fix1
or
collink -v -s -ia -oa -p fix1.icm target.icm
proofer.icm target_proofer_fix1.icm
Note that the above example is a simple one - you should use all
the same options as you used to create your initial proofer.icm
or target_proofer.icm,
with the addition of the "-p fix1.icm" option to specify the
abstract correction profile be applied.
Use the proofer_fix1.icm
or target_proofer_fix1.icm
to print out the test chart again, and read it in, resulting in chart2.ti3 file.
Lets check how well the proofing system matches after this first
round of refinement using verify:
verify reference.ti3 chart2.ti3
>>>
We
can
then start another round of improvement:
We refine our previous abstract correction fix profile using
refine:
refine -v reference.ti3 chart2.ti3
proofer.icm fix1.icm fix2.icm
Applying this new abstract profile to our process for creating the
proofing device profile or link again:
revfix -v -1 -ke -p fix2.icm proofer.icm
proofer_fix2.icm
or
copy proofer.ti3 proofer_fix2.ti3
colprof -v -p fix2.icm proofer_fix2
or
collink -v -s -ia -oa -p fix2.icm target.icm
proofer.icm target_proofer_fix2.icm
Use the proofer_fix2.icm
or target_proofer_fix2.icm
to print out the test chart again, and read it in, resulting in chart3.ti3 file.
Check again how well the proofing system matches after this first
round of refinement using verify:
verify reference.ti3 chart3.ti3
Rounds of improvements can be continues by looping back to >>>, being careful to
increment the names of the fixN.icm,
proofer_fixN.icm or target_proofer_fixN.icm and chartN.ti3files. Stop when
exhausted, or if it looks like things are getting worse, rather
than better. If the latter happens, it might be good to revert to
the results from a previous round.