1.51 to 1.62 transition for RGB to CMYK conversions

[Production]

Graeme

I'm looking to move from v1.51 to v1.62 for device link RGB to CMYK conversions.

I read through the change logs and assumed that there wouldn't be much difference in the resulting separations… but there is a very noticeable difference in dark reds.

The numbers look like 10 per cent of the cyan channel data is getting ported over to yellow.

This produces a CMYK result that is probably closer to a Photoshop conversion than v1.51 but I really like the control and detail you get from keeping as much information on the cyan plate as possible :-)

Can you give me more information about the changes you've made between these two releases?

My setup is simple -- I'm just using collink to make a device link profile from say AdobeRGB(1998) to ISO Coated v2 (300) and then converting using cctiff.

Cheers

[BC Rider]

Question: Does the tone curve of the source space affect the gamut mapping in Colprof? (-S command)?

I edit/print with Adobe Lightroom which uses a non-standard colour space (ProPhoto RGB but with a LINEAR gamma).

I have no icm/icc file for such a space so can't provide it to Colprof. But I can provide a standard ProPhoto RGB profile (gamma 1.8).

[Graeme Gill]

Post by BC Rider
Question: Does the tone curve of the source space affect the gamut mapping in Colprof?
(-S command)?

Hi,
The source color space is an integral
part of creating a gamut mapping - it is what defines the "from" for the mapping.
So the response of the input space determines it's gamut surface shape which in turn
will determine the gamut mapping. But I think that two source profiles that have
identical gamut surfaces will result in the same gamut mapping, since gamut
mapping is done in a PCS derived colorspace.

Post by BC Rider
I have no icm/icc file for such a space so can't provide it to Colprof. But I can
provide a standard ProPhoto RGB profile (gamma 1.8).

Hmm. In theory it should be fine to use that profile to define the source gamut.
It's always possible that I've overlooked something though.

Graeme Gill.

[Alan Goldhammer]

I do all my initial photo editing in Adobe Lightroom and move to Photoshop
for only those things that cannot be accomplished in LR. The Photoshop
color space is set for ProPhoto as that's the one LR uses. I've always soft
proofed with the ProPhoto color space and my ArgyllCMS profiles are prepared
with the standard ProPhoto profile. The visual match between my print and
screen are spot on which is what one wants.

Alan

-----Original Message-----
From: argyllcms-bounce-***@public.gmane.org [mailto:argyllcms-bounce-***@public.gmane.org]
On Behalf Of Graeme Gill
Sent: Thursday, December 19, 2013 2:53 AM
To: argyllcms-***@public.gmane.org
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S

Post by BC Rider
Question: Does the tone curve of the source space affect the gamut mapping in Colprof?
(-S command)?

Hi,
The source color space is an integral
part of creating a gamut mapping - it is what defines the "from" for the
mapping.
So the response of the input space determines it's gamut surface shape which
in turn will determine the gamut mapping. But I think that two source
profiles that have identical gamut surfaces will result in the same gamut
mapping, since gamut mapping is done in a PCS derived colorspace.

Post by BC Rider
I have no icm/icc file for such a space so can't provide it to Colprof.

But I can

Post by BC Rider
provide a standard ProPhoto RGB profile (gamma 1.8).

Hmm. In theory it should be fine to use that profile to define the source
gamut.
It's always possible that I've overlooked something though.

Graeme Gill.

[BC Rider]

----------------------------------------

Post by Alan Goldhammer
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S
Date: Thu, 19 Dec 2013 05:59:06 -0500
I do all my initial photo editing in Adobe Lightroom and move to Photoshop
for only those things that cannot be accomplished in LR. The Photoshop
color space is set for ProPhoto as that's the one LR uses.

Yes, LR will export to Prophoto RGB (or any other color space actually) without problems. But exporting images is something different.

It is misleading to say LR uses Prophoto RGB space though...the chromaticities are the same but the gamma is not. Actually LR uses several different color spaces depending on what it is doing.

Post by Alan Goldhammer
I've always soft proofed with the ProPhoto color space

That seems an unusual workflow as a printer color space is normally used when soft proofing.



----------------------------------------

Post by Alan Goldhammer
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S
Date: Thu, 19 Dec 2013 05:59:06 -0500
I do all my initial photo editing in Adobe Lightroom and move to Photoshop
for only those things that cannot be accomplished in LR. The Photoshop
color space is set for ProPhoto as that's the one LR uses. I've always soft
proofed with the ProPhoto color space and my ArgyllCMS profiles are prepared
with the standard ProPhoto profile. The visual match between my print and
screen are spot on which is what one wants.
Alan
-----Original Message-----
On Behalf Of Graeme Gill
Sent: Thursday, December 19, 2013 2:53 AM
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S

Question: Does the tone curve of the source space affect the gamut

mapping in Colprof?

(-S command)?

Hi,
The source color space is an integral
part of creating a gamut mapping - it is what defines the "from" for the
mapping.
So the response of the input space determines it's gamut surface shape which
in turn will determine the gamut mapping. But I think that two source
profiles that have identical gamut surfaces will result in the same gamut
mapping, since gamut mapping is done in a PCS derived colorspace.

I have no icm/icc file for such a space so can't provide it to Colprof.

But I can

provide a standard ProPhoto RGB profile (gamma 1.8).

Hmm. In theory it should be fine to use that profile to define the source
gamut.
It's always possible that I've overlooked something though.
Graeme Gill.

[BC Rider]

----------------------------------------

Date: Thu, 19 Dec 2013 18:53:03 +1100
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S

Question: Does the tone curve of the source space affect the gamut mapping in Colprof?
(-S command)?

Hi,
The source color space is an integral
part of creating a gamut mapping - it is what defines the "from" for the mapping.
So the response of the input space determines it's gamut surface shape which in turn
will determine the gamut mapping. But I think that two source profiles that have
identical gamut surfaces will result in the same gamut mapping, since gamut
mapping is done in a PCS derived colorspace.

I have no icm/icc file for such a space so can't provide it to Colprof. But I can
provide a standard ProPhoto RGB profile (gamma 1.8).

Hmm. In theory it should be fine to use that profile to define the source gamut.
It's always possible that I've overlooked something though.
Graeme Gill.

Thanks Graeme. Lightroom is quite opaque in its internal handling of color spaces so some experimenting will be required anyway.

But thinking a bit further, won't such a large source space tend to desaturate colors when printed? The ProPhoto RGB color gamut is larger than my printer and most of my images!

Is it reasonable to consider using AdobeRGB as the source gamut assumption? If I understand correctly, colors in excess of AdobeRGB will then clip but the rest will be scaled to my printer gamut producing better overall saturation on typical images.

If this isn't crazy talk, I suppose one could (and should) make two profiles and use which ever one looks best for a given image.

[Iliah Borg]

Post by Graeme Gill

I have no icm/icc file for such a space so can't provide it to Colprof. But I can
provide a standard ProPhoto RGB profile (gamma 1.8).

Hmm. In theory it should be fine to use that profile to define the source gamut.
It's always possible that I've overlooked something though.

As far as I remember recent Lightroom allows for any output colour space, so maybe ProPhoto gamma 1 output is a better input space for correction profiles.

--
Best regards,
Iliah Borg
LibRaw, LLC

[Alan Goldhammer]

I didn't reply to your earlier post but will here. It's not opaque at all.
LR uses Melissa RGB which is the ProPhoto gamut but with a linear gamma.
See Jeff Schewe's comment here: http://forums.adobe.com/message/5360957

-----Original Message-----
From: argyllcms-bounce-***@public.gmane.org [mailto:argyllcms-bounce-***@public.gmane.org]
On Behalf Of BC Rider
Sent: Thursday, December 19, 2013 2:30 PM
To: argyllcms-***@public.gmane.org
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S

Thanks Graeme. Lightroom is quite opaque in its internal handling of color
spaces so some experimenting will be required anyway.

[BC Rider]

----------------------------------------

Post by Alan Goldhammer
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S
Date: Thu, 19 Dec 2013 16:41:44 -0500
LR uses Melissa RGB which is the ProPhoto gamut but with a linear gamma.

No, Melissa RGB is not linear gamma. It uses an sRGB like tone curve.
Melissa RGB is used to calculate the histogram display and RGB readout but is not used for editing. The actual editing/processing space in LR is ProPhoto RGB colors with a linear gamma.

Post by Alan Goldhammer
See Jeff Schewe's comment here: http://forums.adobe.com/message/5360957

Yes, Schewe has it correct. You may want to re-read his comment.

Post by Alan Goldhammer
It's not opaque at all.

I only say that based on how many people are confused by it...

As I said earlier LR used many different spaces. The Library module displays previews in AdobeRGB space, the Develop modules displays previews in ProPhoto gamut with a sRGB tone curve applied, the Slideshow modules uses AdobeRGB, and the Web module previews are in sRGB. Untagged input images (non-RAW) are assumed to be sRGB and converted to LR processing space which is ProPhoto gamut with linear gamma. And so it goes, given all that who really knows what space Adobe decided to use for the Print module! None of this is visible to the user and in any case none of it can be changed.

Anyway, this list is about Argyll not Lightroom. Suffice to say, creating the optimum source space when editing and printing in Lightroom is not so obvious to me hence the other posts in this thread.

[Graeme Gill]

Post by BC Rider
But thinking a bit further, won't such a large source space tend to desaturate colors
when printed? The ProPhoto RGB color gamut is larger than my printer and most of my
images!

Yes, if you assume the source gamut is the boundary of the color space (or any wide
gamut color space), you risk over compression and de-saturation.

That's why you need to use a different workflow for these situations, and why
you can specify an image gamut in colprof and collink. The source image gamut
defines the volume you wish to preserve in the gamut mapping. You could use
the gamut of a smaller colorspace such as sRGB or AdobeRGB using iccgamut,
or you could create an image specific gamut from one or more images using
tiffgamut: <http://www.argyllcms.com/doc/tiffgamut.html>.

Ideally, determining the source gamut to be preserved is part of the creative
endeavor, something you might hand tune for each image.

Graeme Gill.

[BC Rider]

----------------------------------------

Date: Fri, 20 Dec 2013 13:54:43 +1100
Subject: [argyllcms] Re: Adobe Lightroom and Colprof -S

Post by BC Rider
But thinking a bit further, won't such a large source space tend to desaturate colors
when printed? The ProPhoto RGB color gamut is larger than my printer and most of my
images!

Yes, if you assume the source gamut is the boundary of the color space (or any wide
gamut color space), you risk over compression and de-saturation.
That's why you need to use a different workflow for these situations, and why
you can specify an image gamut in colprof and collink. The source image gamut
defines the volume you wish to preserve in the gamut mapping. You could use
the gamut of a smaller colorspace such as sRGB or AdobeRGB using iccgamut,
or you could create an image specific gamut from one or more images using
tiffgamut: <http://www.argyllcms.com/doc/tiffgamut.html>.

I see...thanks a lot!

Presumably then "Colprof -S Adobe1998.icm", and "Colprof -S Prophoto.icm -g Adobe1998.gam" are the same? If so, is there any reason to choose one over the other?

Ideally, determining the source gamut to be preserved is part of the creative
endeavor, something you might hand tune for each image.

Is there a general way to expand/contract the gamut when using iccgamut/tiffgamut? I didn't see one. Seems that would be useful. (I did see that tiffgamut can contract the gamut by filtering colors in the image.)

[Graeme Gill]

Post by BC Rider
Is there a general way to expand/contract the gamut when using iccgamut/tiffgamut? I
didn't see one. Seems that would be useful.
(I did see that tiffgamut can contract
the gamut by filtering colors in the image.)

Yes, that's it.

Graeme Gill.

[BC Rider]

Post by Graeme Gill

I have no icm/icc file for such a space so can't provide it to Colprof. But I can
provide a standard ProPhoto RGB profile (gamma 1.8).

Hmm. In theory it should be fine to use that profile to define the source gamut.
It's always possible that I've overlooked something though.

Hi Graeme,

I tried to test this and would appreciate your (or anyone's) input on what I've done and the results.

I discovered my image editor (Photoline) will easily export a standard ProPhotoRGB profile as a linear gamma profile so I now have both profiles (gamma 1.8 and gamma 1.0).

I then generated two identical printer profiles changing only the -S source space:

E.g. colprof -v -y -S ProPhoto.icm -cmt -dpp -O Test_PrinterProphotoSrc.icm plainpaper.ti3, and
colprof -v -y -S ProPhoto_lineargamma.icm -cmt -dpp -O Test_PrinterProphotoSrc_lineargamma.icm plainpaper.ti3

I couldn't find a way in Argyll to compare two profiles (especially with perceptual intents) so I used xicclu and generated ten neutral LAB points from 0-100 as input and compared the RGB results:

E.g. xicclu -fb -ip -s 100 test_printerprophotosrc.icm
xicclu -fb -ip -s 100 test_printerprophotosrc_lineargamma.icm


Conversion results using the standard ProPhotoRGB profile as source are:
0.000000 0.000000 0.000000 [Lab] -> Lut -> 0.000000 0.000001 0.000001 [RGB]
10.000000 0.000000 0.000000 [Lab] -> Lut -> 1.834441 0.577441 1.352944 [RGB]
20.000000 0.000000 0.000000 [Lab] -> Lut -> 4.969248 3.633593 3.668262 [RGB]
30.000000 0.000000 0.000000 [Lab] -> Lut -> 10.861680 9.784255 7.057493 [RGB]
40.000000 0.000000 0.000000 [Lab] -> Lut -> 18.199058 19.228959 14.434314 [RGB]
50.000000 0.000000 0.000000 [Lab] -> Lut -> 27.608013 31.531528 25.658035 [RGB]
60.000000 0.000000 0.000000 [Lab] -> Lut -> 44.758474 49.856496 41.119166 [RGB]
70.000000 0.000000 0.000000 [Lab] -> Lut -> 62.296613 65.503135 59.485163 [RGB]
80.000000 0.000000 0.000000 [Lab] -> Lut -> 75.861582 76.318738 75.158597 [RGB]
90.000000 0.000000 0.000000 [Lab] -> Lut -> 90.325365 88.927061 90.362753 [RGB]
100.000000 0.000000 0.000000 [Lab] -> Lut -> 99.998916 99.998529 99.998888 [RGB]

Conversion results using the linear gamma ProPhoto profile as source are:
0.000000 0.000000 0.000000 [Lab] -> Lut -> 0.000000 0.000001 0.000001 [RGB]
10.000000 0.000000 0.000000 [Lab] -> Lut -> 1.736744 0.912101 1.447588 [RGB]
20.000000 0.000000 0.000000 [Lab] -> Lut -> 4.739884 3.858732 3.784321 [RGB]
30.000000 0.000000 0.000000 [Lab] -> Lut -> 10.502501 9.830148 7.404452 [RGB]
40.000000 0.000000 0.000000 [Lab] -> Lut -> 18.085888 19.168456 14.772755 [RGB]
50.000000 0.000000 0.000000 [Lab] -> Lut -> 27.668562 31.410477 25.691384 [RGB]
60.000000 0.000000 0.000000 [Lab] -> Lut -> 44.787028 49.709416 41.132378 [RGB]
70.000000 0.000000 0.000000 [Lab] -> Lut -> 62.315907 65.377592 59.449139 [RGB]
80.000000 0.000000 0.000000 [Lab] -> Lut -> 75.898455 76.170333 75.126697 [RGB]
90.000000 0.000000 0.000000 [Lab] -> Lut -> 90.308257 88.752226 90.211267 [RGB]
100.000000 0.000000 0.000000 [Lab] -> Lut -> 99.998896 99.998516 99.998868 [RGB]

My observations:
1) Results are NOT identical
2) but results are generally close (not noticeable?)
3) except near black where there is a noticeable difference

Gamutvision can compare two profiles with each other and show the differences (dE) so I tried that also. The driving function was a little different, along the sRGB gamut boundary rather than along the grey axis, but results appear similar. I've attached a file showing the Gamutvision screen.

(Tentative) Conclusion: Colprof is affected by gamma of the source space, especially towards black.

I would appreciate any comments. Thanks.

[Graeme Gill]

Post by BC Rider
(Tentative) Conclusion: Colprof is affected by gamma of the source space, especially towards black.

It's hard to know exactly why this is without a very detailed investigation. The gamut
suface detail level is influenced somewhat by the profile response, because gamut surface
test points are generated in device space. So the linear gamma source gamut will have
slightly less detail (curvature) near black. Whether that fully explains the difference,
I'm not sure.

Graeme Gill.

[Graeme Gill]

Post by Production
The numbers look like 10 per cent of the cyan channel data is getting ported over to
yellow.
Can you give me more information about the changes you've made between these two releases?

Hi,
it's almost impossible to determine from such a high level description
of one particular device & situation.

The only practical way of investigating this type of thing is to have an actual
example to compare - ie. .ti3 file, command line, example source image.

Graeme Gill.

[Production]

Post by Graeme Gill
it's almost impossible to determine from such a high level description
of one particular device & situation.

Sorry :)

I'm creating the device links from AdobeRGB to ISO Coated v2 (300) using -ip -ke -cmt -dpp

The results on proof are for the most part favourable, but also clearly noticeable -- which I wasn't expecting.

Post by Graeme Gill
The only practical way of investigating this type of thing is to have an actual
example to compare - ie. .ti3 file, command line, example source image.

I'll send you a small section of an ad image, the resulting CMYKs and the device links too off-list (via WeTransfer).

Happy to share the numbers on-list but can't do that with the image unfortunately.

Cheers