ICM Profile comparison - ArgyllCMS - i1Profiler

[Ink stained Fingers]

I'm continuing here a specific subject originally raised in the ArgyllCMS thread how profiles compare created with different software - ArgyllCMS vs. i1Profiler with various patch field counts.
This thread connects to the other thread starting this slubject in postings # 427 - 475

https://www.printerknowledge.com/th...s-profiling-on-your-computer.8570/post-136740

One specific question got discussed which rendering intent assures most accurate color reproduction - rel. colorimetric or absolute colorimetric.
I'm continuing this subject here since it goes beyond the original subject of the ArgyllCMS thread. I have partially restarted the testing and use a small patch sheet with 14 gray fields - it's easy with measuring gray and to see whether they remain neutral or not - the a and b Lab values directly indicate any tint in the gray. I got a profile created for my L1800 with Netbit paper by @pharmacist , and I'm comparing this against a profile I created with the XRite i1Profiler package - with 2880 patches to get most accuracy - hopefully.
I printed the gray patches with both profiles - with the rel. col. rendering intent; the abs. col. rendering intent showed significant deviations in the previous tests.

There are 14 gray patches - incl. white and black with the RGB values listed in the left column and the corresponding L values - and b remain 0 for neutral grays.

I scanned these patches with these results:

It is interesting to see that the gray axis with the ArgyllCMS profile does not deviate that much from the neutral axis as the values for the i1 Profiler profile do - there is a pretty strong overall tint into a blue white direction.
The gray patch sheet is printed with the rel. col. rendering intent with both profile options; the i1Profiler software offers an option 'Gray neutralisation' for the profile creation process, I turned up this option - it does not bring the gray values closer to neutral as the ArgyllCMS profile does. I don't know at this time which parameter is causing these deviations.

 

[Ink stained Fingers]

I did some more tests attempting to print neutral grays; I created several profiles with i1Profiler - based on patch sheets with 96 or 283 or 450 or 2880 patches, I printed a wedge with 14 gray values incl. black and white and measured the resp gray values. I printed both with the rel. col. or perceptual rendering intent - with disappointing results - no prints show neutral values on most of the patches - not at the white and black point, the b value - direction blue - is maeandering around similarly to the numbers above, there is one exception - the profile based on just 96 patches performs the best - slightly better than the ArgyllCMS profile by @pharmacist , I assume there are only a few control points along the gray axis so that the software is interpolating across a wider range. It's all disappoting at this time that a regular profile does not print accurately neutral grays. The XRite sofware with the ColorMunki, I1STudio or i1Profiler allows an iterating optimization step - I'll try this as the next action if this changes /improves the gray printing at all.
The XRite software does not allow any further profile evaluation or editing, other profiling software like EFI or CoPra by ColorLogic offer functions like profile editing and adjustment - I don't know how and if that is working and supplying the expected results.

 

[Ink stained Fingers]

I found and article here that technically neutral grays - with a and b Lab values of 0 - are visually not considered as neutral - it should have a slight blue cast - with a b=- 1 . I don't know if that is directly visible making such a difference, the article goes into some detail of this:

'Technically, midtone neutral in L*a*b* is L*50, a*0, b*0. However, visually we do not
accept this. In my opinion, we have come to think that neutral gray must be
slightly cool because it is the ugly step-sister of blue. I’m fine with that as long
as it is documented as a specification I can point to and know when I have
attained it.
True visual midtone neutral gray is documented by GRACoL 7 and speci-
fied by G7 as L*59, a*0, b*-1.'

'Neutral gray is technically L*50, a*0, b*0, but we don’t accept this color visu-
ally. We tend to see neutral as having a slight blue cast. Therefore, neutral
gray is specified slightly blue '

Copied from here on page 4

https://www.nazdar.com/Portals/0/NewsItems/Articles/ThePowerOfNeutralGray-MikeRuff.pdf


Currently most of my gray patches deviate much more - larger -b- values - wider than b=-1 even if the profiling software would follow the above rule.

 

[nertog]

I found and article here that technically neutral grays - with a and b Lab values of 0 - are visually not considered as neutral - it should have a slight blue cast - with a b=- 1 . I don't know if that is directly visible making such a difference, the article goes into some detail of this:

'Technically, midtone neutral in L*a*b* is L*50, a*0, b*0. However, visually we do not
accept this. In my opinion, we have come to think that neutral gray must be
slightly cool because it is the ugly step-sister of blue. I’m fine with that as long
as it is documented as a specification I can point to and know when I have
attained it.
True visual midtone neutral gray is documented by GRACoL 7 and speci-
fied by G7 as L*59, a*0, b*-1.'

'Neutral gray is technically L*50, a*0, b*0, but we don’t accept this color visu-
ally. We tend to see neutral as having a slight blue cast. Therefore, neutral
gray is specified slightly blue '

Copied from here on page 4

https://www.nazdar.com/Portals/0/NewsItems/Articles/ThePowerOfNeutralGray-MikeRuff.pdf


Currently most of my gray patches deviate much more - larger -b- values - wider than b=-1 even if the profiling software would follow the above rule.

What is perceived as neutral grey depends a lot on the chromatic adaptation state of our visual system. Simply put: our eye + brain transforms paper white to a neutral tint, so the greyscale has to be shifted to follow that paper white point. This is exactly what the relative col. intent is doing.
There is a ton of research on chromatic adaptation and visual perception of neutral tones. The above white paper offers an opinion of one guy..and I don´t agree with his statement The quoted GRACol document lists L*59, a*0, b*-1 as perceptually neutral, and that measurement is slightly yellower than their paper white of L*95, a*0, b*-2 that is used as a base.

@ISF: I can send you my standard set of calibration patches (16-bit TIFF) for you to print out and send back to me. I´ll measure them here and send you the Argyll-based ICC-profile back for testing. With my printer, those ICC-profiles result in a relatively neutral greyscale and correct relative vs. absolute colorimetric behavior. If you send me your profile I can also have a look at the predicted neutral axes for the different rendering intents.

 

[Ink stained Fingers]

There is a ton of research on chromatic adaptation and visual perception of neutral tones. The above white paper offers an opinion of one guy..and I don´t agree with his statement

Yes, that's just one comment, and I want to get around all that research and just measure grays as neutral grays to start with and if and which differences exist with different profiles - ARgyllCMS vs. i1Profiler - low or high patch count, and I'll try to get hold of an i1Studio/ColorMunki unit. As mentioned I'll try to test next if the optimization/iteration mode makes any difference and makes an improvement.
And yes - please send me your calibration patches - I'll continue to use - for reasons of comparison with the previous tests - the L1800 to print on the Netbit/Aldi paper. Which spectro are you using ?

 

[Ink stained Fingers]

I found an interesting article - already almost 20 years old - dealing exactly with the effect that an ICC-profile does not render grays neutral in print. The complete article needs to be requested for download by the author , it describes a method how to compensate these deviations and to correct the resp. table in the ICC profile. But this makes it impossible for a regular creator and user of ICC -profiles since you need access to the profile tables with an editor and exact knowledge of the table layout.

https://www.researchgate.net/publication/253775517_Neutral_gray_adjustment_in_printer_ICC_profiles

 

[nertog]

Which spectro are you using ?

I´m using a relatively old I1 Pro rev. D spectro.

 

[Ink stained Fingers]

With or without an UV filter - 'UV Cut' would be marked on the unit label

 

[nertog]

With or without an UV filter - 'UV Cut' would be marked on the unit label

Standard, not UV-cut

 

[Ink stained Fingers]

There is a wast mount of postings on the internet about B/W printing, neutral grays etc but only a veery few of these postings actually look to the spectral composition of grays and the Lab values - one of these few posting is this one

https://www.nazdar.com/Portals/0/NewsItems/Articles/ThePowerOfNeutralGray-MikeRuff.pdf

commenting as well that a gray viewed as 'neutral' actually has a small blue tint. I take it as such.

I did a test with a i1Studio unit, the successor of the ColorMunki, I created a profile for a budget glossy paper, the profiling process starts with a patch sheet with 50 patches and an optimzation cycle with another 100 patches on 2 more sheets. i1Studio has a specific flow for B/W profiles which are actually regular color profiles but with more patch colors along the gray axis. Threre is an option for the generation of the profile how the grays should be tuned - normal - sepia or cool, I selected 'normal' The i1Studio spectro by XRite uses an UV filter, so any UV related properties of the paper are not measured and processed into the profile. The paper used for this test comes with quite a high amount of optical brighteners. The overall gamut of this profile is very much identical with a profile I created earlier for this paper with the i1Profiler package - as well not using UV data - the M2 measuring mode.

I printed a little gray step wedge with 14 gray patches and measured the Lab values, I have removed in this table the actual L-luminance values for 2 scans - one in the M0-UV mode and another scan in the M2 mode, and I removed the a-values as well - they meander rather close between -1,5 to +1,5 and are of no concern.

The column for the b-values - show the impact of the optical brighteners - the white point drops into the blue direction with -b= -8,68 vs. -0,86 for a viewing light w/o UV content. That's a pretty strong and visible difference and shows a principal problem beyond gray levels - you cannot profile a paper at the same time for 2 different illumination conditions when OBA's are present.

The black point with a b~-5 has a pretty cool touch which is typical for dye inks, pigment inks are closer to a neutral value. The numbers above further show how the gray axis is placed between the white and the black point - there are about nowhere gray neutral values with a b-value close to zero.