[argyllcms] Re: Comparing Input / Output ICC profiles the correct way?
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Thu, 14 Oct 2010 19:02:23 +1100
gklima@xxxxxxxxxxxxxx wrote:
I'm trying to compare input with output profiles as well as output with
output profiles. These are all 4 channel CMYK profiles.
Case 1 - Input with Output profile:
Input profile: IsoCoated Euroscale ECI v2
Output profile: Ink-jet printer profile generated with Caldera (iPrism
v8)
Goal: the correct intersecting volume as calculated with viewgam -i
Question: What rendering intent is the correct one for such a comparison
(absolute RI?)? I want to know how much (in %) of the input profile is
printable with the output profile. And should I use iccgamut with "-fb" or
"-ff"? What and why is the correct setting.
This depends on the rendering intent you intent do use in reproducing
IsoCoated Euroscale ECI v2 using the Ink-jet. If your goal is proofing,
then absolute intent is usually desirable. For the source profile (Euro ECI v2),
you want to use -ff, since this selects the A2B table that will be used
in linking it. For the destination profile (Ink-jet), you probably want
to use -fb if your goal is to figure out what actual gamut will be
reproduced, since this selects the B2A table that will be used in
linking using a typical passive/dumb CMM.
If you wanted to know what gamut could possibly be reproduced, then
selecting the A2B table of the destination profile (-ff) and the
ink limit, would be the right choice.
Case 2 - Output with Output profile:
Output profile: Ink-jet printer profile generated with Caldera (iPrism
v8)
Output profile: Ink-jet printer profile generated with Caldera (iPrism
v8)
Goal: the correct intersecting volume as calculated with viewgam -i
Question: What rendering intent is the correct one for such a comparison
(absolute RI?)? I want to know how much (in %) of one profile is included
in the other. And should I use iccgamut with "-fb" or "-ff"? What and why
is the correct setting.
Again, this depends on what you are trying to do, although to a lesser extent.
If (for instance) the two profiles used the same paper, then essentially
there will be no difference between absolute and relative colorimetric intents.
If the paper is different (ie. different white point), then it will make a
difference, and so the selection will depend on what you are trying to do.
If you are interested in using the profiles for proofing, the absolute
colorimetric will typically be appropriate. If you are using them for producing
pleasing images, and therefore would select a relative colorimetric type
intent during linking (match white points), then use that intent for
gamut comparison. Once again, -fb will show you what the B2A table can
actually produce, while -ff with an ink limit will show you what the device
is capable of. They may be different, depending on the capabilities of
the profile maker in creating the B2A table.
Another Scenario:
Assume I did have two printers with two ink sets which print the same
profiling data on the same media (same white point) which is then profiled
by the same spectrophotometer. Comparing both CMYK ink sets (in terms of
printable gamut) to see with which printer you can print "more colors".
In this case it seems clear: absolute photometric.
If the white point is the same, it won't make any difference whether its
absolute or relative colorimetric.
But comparing an artificial input profile as for example Iso Coated ECI v2
which has a default RI of perceptive (what I clearly not understand, how a
single profile could have a RI. I though only the relation / translation
between two profiles could have a RI) with a printers (output) profile in
terms of printable gamut with ideally (theoretically) delta E equal to
ZERO (not possible in practice, I know!!).
A default is just that, a default. It can (usually is) overridden. A single
profile can have an intent, because of the way the ICC organised things.
cLUT profiles have three forward tables (A2B, device->PCS) and three backwards
tables (B2A, PCS->device), each table representing a possible intent
(colorimetric,
perceptual & saturation). A fourth intent is created from colorimetric by
re-forming the absolute data using the white point tag.
With a passive/dumb CMM therefore, it's possible to select a source
intent and a destination intent. In practice this approach to dealing
with linking has serious drawbacks, and is not capable of general
gamut mapping (see <http://www.argyllcms.com/doc/iccgamutmapping.html>).
To do general gamut mapping, an active/smart CMM is needed, and it will
have an overall intent selection (See collink -g and -G for an example of
an active/smart linker).
Assuming exactly the same media as used for the isocoated with also the
same spectrophotometer then it would too be absolute RI. But using a
different media with another white point then is should be relative RI?
It depends on how you intend to link it. If you are doing proofing, and
want to exactly reproduce colors colorimetrically, including reproducing
the paper color, then you will want to use absolute colorimetric. If you
want to produce pleasing images, then you will use some white point
adaptive intent such as relative colorimetric, perceptual or saturation.
What seems clear is that the perceptive and saturation RIs are not usable
for a gamut comparision of two profiles.
They are if you wish to examine what gamut these B2A tables are
capable of.
Maybe someone could explain this to me please or point me to the right
documentation.
A lot of this is implicit in the ICC specification.
1) Why does a profile itself have a RI (a RI in relation to what? To the
"unlimited" Lab PCS?)
It hints at which intent A2B and B2A table to use by default.
2) how to compare gamuts of profiles under what circumstances concerning
delta E == ZERO printable colours (ideally, theoretically) with which
rendering intent (assuming a most neutral comparison for the most general
case, not assuming any special print job, just to say with this printer I
can print X% of the gamut in relation)
You can't. A real print job will use a specific set of conditions,
and this will affect what range of colors can be reproduced.
You can of course make a specific choice, and hope that it is indicative
of other situations.
Maybe I'm just asking the wrong (impossible scenario) question!?
But there should (at least I hope) be a standard for comparing presentable
colours of different devices!?
It depends on the intended use. For proofing one would initially look
at the absolute colorimetric gamut of the devices (-ff with an ink limit).
For other uses where relative colorimetric, perceptual or saturation style
intents will be used, one would look at the relative colorimetric gamut
of the devices (-ff with an ink limit). If one is evaluating how well
a particular profiler has used the available gamut in its B2A tables,
one would select -fb of a destination profile.
Graeme Gill.
Other related posts:
