Validating Printer Profiles

[CowZ]

Hi all,

after being helped so well with a printing software, I have a next question / followup. I am now investigating how to best profile my printer. I do this more out of fun and theoretical interest than that I hope for real-life visible better image quality. That is to say, I'm rather happy with the default i1Studio Profiler software and results. I expect though, that it may be possible to be better or with less paper. So, now I'm spending paper after paper to improve my profiling workflow using Argyll ;-).

After now having multiple different ways to profile my printer/paper (i1Studio proprietary, and various settings for Argyll), I have problems to decide which workflow is best. I can, of course, print test images and compare them visually, but this is too subjective to me. I would like to find an objective measure.

I am rather sure, that this is possible: I profile my printer, so that he reproduces the colors in the most accurat way. So, with a perfect printer and a perfect ICC profile and perfect paper, I should be able to read the color values with my i1Studio and they should be the same as the values I set in my image creation tool.

But: I get *very* high deltaE values (around Mean 10dE, Max 23dE) and I was hoping for less - even given my poor-man's hardware (i1Studio and rather cheap photoprinter epson ET-7750).

My current workflow: I create a patch set with targen/printtarg and print it with my ICC profile in place. I am expecting that the software&printer is now trying to output the same Lab-Values that are in the created tiff. I use relative colorimetric rendering intent. I also tried to reproduce the x-rite ColorChecker in Gimp and then print the resulting tiff.

I also worry a bit, as I cannot find anything useful on this topic at all.

Kind regards,

Lasse

PS: I posted a similar question to the Argyll mailinglist, but responds were rare/none and I do also not want to limit me to Argyll. Theoretical approaches may also help, as well as usage of other software.

 

[palombian]

I tried once to make a photo from the colorchecker card, import it in LR with a camera profile, and print it with the appropriate printer profile.

In theory the colors of the print should be the same as the original.

I measured the Lab values with the Colormunki and the Colorpicker app (part of the original software, don't know if it works with the i1Studio).

The differences were several Delta E points. Not sure if I used the Colorchecker Camera Calibration correctly.

I gave up this silly idea since it was not important to me.

While the accuracy of the Colormunki is not on par with more sophisticated devices, the the variation is <=1.0 Delta E, so not visible with the eye, sufficient for amateur use.

 

[technoid]

Hi all,

after being helped so well with a printing software, I have a next question / followup. I am now investigating how to best profile my printer. I do this more out of fun and theoretical interest than that I hope for real-life visible better image quality. That is to say, I'm rather happy with the default i1Studio Profiler software and results. I expect though, that it may be possible to be better or with less paper. So, now I'm spending paper after paper to improve my profiling workflow using Argyll ;-).

After now having multiple different ways to profile my printer/paper (i1Studio proprietary, and various settings for Argyll), I have problems to decide which workflow is best. I can, of course, print test images and compare them visually, but this is too subjective to me. I would like to find an objective measure.

I am rather sure, that this is possible: I profile my printer, so that he reproduces the colors in the most accurat way. So, with a perfect printer and a perfect ICC profile and perfect paper, I should be able to read the color values with my i1Studio and they should be the same as the values I set in my image creation tool.

But: I get *very* high deltaE values (around Mean 10dE, Max 23dE) and I was hoping for less - even given my poor-man's hardware (i1Studio and rather cheap photoprinter epson ET-7750).

My current workflow: I create a patch set with targen/printtarg and print it with my ICC profile in place. I am expecting that the software&printer is now trying to output the same Lab-Values that are in the created tiff. I use relative colorimetric rendering intent. I also tried to reproduce the x-rite ColorChecker in Gimp and then print the resulting tiff.

When printing specific colors you must use Absolute Colorimetry

See this collection of Lab colors. These are in steps of 10 for a* and b* with each image in steps of 10 L*. You can print this in Photoshop selecting Abs. Col. and Photoshop manages color. You might have to scale them up to be able to measure spot color.

Profile Evaluation Images

Also interesting to see what gets printed that are out of gamut by soft proofing these against your printer profile.

I also worry a bit, as I cannot find anything useful on this topic at all.

Kind regards,

Lasse

PS: I posted a similar question to the Argyll mailinglist, but responds were rare/none and I do also not want to limit me to Argyll. Theoretical approaches may also help, as well as usage of other software.

 

[Ken60]

Firstly might I suggest you post this query in the Printing forum where a few experienced printers may offer best advice. Look for pixelgenius, Mark McCormick , or Keith Cooper.

I have to suggest that your printer is a 3 colour + black machine , and as such it relies on dot dithering for lighter shades. Think of more white paper than ink for the lightest tones. Many difficult to produce colours are difficult enough to create with the 11 ink printer I am using , and some manufacturers add Red, Violet, Green and Orange as well as light versions of the cmyk basics. When you start to become fastidious about perfect colour I would query the chance of perfection with what is basically designed as an " office" or " home" printer. Some of the very top end printers are certified to reach 99% pantone , and can be used as proofing machines at the output end of a design studio . The Epson SC P5000CE is such a machine , and it costs around £2500

 

[Harvey M43]

Hi-

There’s multiple aspects to profile validation. Anders Torger has a good tutorial for the ColorMunki in his ‘Inkjet printer profiling with Argyll and Colormunki’ that includes some hints.

Argyll's Profcheck allows you to run either the creator patchset (best case result) or a different patchset (more realistic?) through the profile to estimate it’s accuracy (delta E) for each patch. It notes the max delta E and average delta E. I use the values in a spreadsheet to generate various other metrics, like the average delta E for the top 90%, or percent of patches with a delta E lower than 1.0 and a graph of the patch's delta E. While ‘fun’ they don’t really tell you more than the profcheck summary values.

Profcheck also has the -w option which creates a graphic (opened by most web browsers) of the delta E as a colored line the length being the delta E and position the color/ color displacement. Besides showing misreads it’s also useful in showing areas where there are problems. In one case I noticed a strange corkscrew, zooming in showed the patches around the dark neutrals were problematic.

As for printing out a target here's one way of doing it:

Print out BabelColor's Tif of the colorchecker in Absolute colorimetric rendering. You can download their Lab values from the same site.

Spot read the patches and compare to the published values.

While interesting really the thing that counts is how the profiles do on your prints. If you can't see it in your print it's valueless...

 

[CowZ]

Thank you, a new set of test images for me

 

[CowZ]

Hi Ken,

thank you for your answer. I thought this was in the "Printers and Printing" forum, is there a better alternative?

Also, I am well aware that my small printer is far from perfect, but my question would stand as it it even with a bigger, better, more expensive printer. And should really be read independent of the hardware I have.

Kind regards, Lasse

 

[CowZ]

Hi Harvey,

so far I thought that profcheck may only be used to compare the ICC with the .ti3 that was used for profile generation. This would then not reflect the actual correctness of the profile, though. I would rather read a new test-patch set with the profile applied to check it's correctness. But if I understand you correctly, this should be possible?

Workflow:

1. Generate profile as previously done
2. use targen + printtarg to create a new testpatch page(s).
3. print, this time *with* the profile in the chain (i.e. using cctiff or simply selecting the profile in the printing software) and absolute colorimetric*
4. use chartread to create a new .ti3 file
5. use profcheck to compare .ti3 file with the ICC profile

Is this going to work? I will see in 30minutes when the testchart has somewhat dried.

Kind regards,

Lasse

*) I thought relative may be better as it should compensate the whitepoint of the paper?

 

[Ken60]

Hi you are in the right place, I was reading the PC forum and a red flag drew me to a thread ..... seems it changed forum to follow the thread. I apologise for the disorientation of my post.

 

[Harvey M43]

Hi Harvey,

so far I thought that profcheck may only be used to compare the ICC with the .ti3 that was used for profile generation. This would then not reflect the actual correctness of the profile, though. I would rather read a new test-patch set with the profile applied to check it's correctness. But if I understand you correctly, this should be possible?

Workflow:

1. Generate profile as previously done
2. use targen + printtarg to create a new testpatch page(s).
3. print, this time *with* the profile in the chain (i.e. using cctiff or simply selecting the profile in the printing software) and absolute colorimetric*
4. use chartread to create a new .ti3 file
5. use profcheck to compare .ti3 file with the ICC profile

Is this going to work? I will see in 30minutes when the testchart has somewhat dried.

Kind regards,

Lasse

*) I thought relative may be better as it should compensate the whitepoint of the paper?

Hi Lasse:

Profcheck with work with any ICC profile (I've used Canon, i1Studio and Argyll ICC profiles) and any .ti3 file (but of course it should be on the same paper for meaningful results).

The reason to use the Colorchecker target is it's patches generally fall within the color gamut of what printers can reproduce on matte paper. If your patches fall outside the printer/paper color gamut you'll get a poor result no matter how good the profile.

I believe Argyll creates white patches in all it's profiles, and I believe uses that, so adding whitepoint compensation in the rendering would double correct.

 

[CowZ]

Hi Harvey,

I can come back with some numbers now: I tried printing absolute colorimetric, but it outputs a light gray where it should be white (so even the non-patch areas at the outside of the tif are gray).

Independent of if I use relative colorimetric with cctiff or select that in Epson Print Layout, I end up with dE values max ~33, avg 12.6.

To me, this looks reasonable high?

Here's details on my calls to ArgyllCMS:

targen -v -d2 -G -e2 -B 2 -g64 -f210 validate
printtarg -v -i CM -h -T300 -pA4 validate
cctiff -ia .\sRGB.icm .\i1_1xa4.icm .\validate.tif validate_abscol.tif
(or use -r for relative colorimetric)
chartread -v -T "0.4" validate
profcheck -v -k -s -I r .\validate.ti3 .\printerprofile.icm

Are those dE values okay or too high? What am I doing wrong? Is the assumption, that the results from printtarg are in sRGB wrong?

How can I create a patch field (readable with chartread + i1Studio) that uses the color checker values?

 

[Harvey M43]

Hi-

I was modifying my answer when you replied (locking my post) here's my revised post-

Profcheck with work with any ICC profile (I've used Canon, i1Studio and Argyll ICC profiles) and any .ti3 file (but of course it should be on the same paper for meaningful results). It simply feeds the patch values printed through the profile and compares that to the measured ones. It’s is intended to use patchsets printed without color management (as targets). Which is how I’ve used it. I’m not sure what the results would be when a profile is used in the printing.

The reason to use the Colorchecker target is it's patches generally fall within the color gamut of what printers can reproduce on matte paper. If your patches fall outside the printer/paper color gamut you'll get a poor result no matter how good the profile. These print values would spot read and compared to their creator (input) values by software (there’s free spreadsheet templates on the web to calculate delta E)

I believe Argyll creates white patches in all it's profiles, and I believe uses that, so adding whitepoint compensation in the rendering would double correct.

I think given that your new tif has a grey cast over the entire sheet means you should step back from profile validation & work out if the grey cast affects all your prints, and what is causing that problem. My personal feeling is any profile giving a double digit error isn't useable.

 

[technoid]

Hi Harvey,

I can come back with some numbers now: I tried printing absolute colorimetric, but it outputs a light gray where it should be white (so even the non-patch areas at the outside of the tif are gray).

There is nothing wrong. This is exactly what you should see.

Here's the problem. Absolute Intent accurately prints only colors that are in gamut. For those colors that are out of gamut it prints the "closest" color in a geometric sense.

White, ie: RGB 255,255,255 is the same as Lab 100,0,0 and it can't be printed because no printable paper is "perfect." Unprinted Photo paper typically doesn't reflect 100% of the light that hits it but only about 85% which roughly corresponds to L*=95 with a* and b* shifted slightly depending on how evenly the paper reflects light of different wavelengths.

What gets printed is the closest color to white in the printable gamut. So a very light sheen is what's printed. Perceptual and Relative scale the printed colors to the paper's actual white. This is what you want printing photos but not when trying to determine profile "accuracy". For that you have to print Absolute and make sure colors are in gamut.

Ctrl-Y in Photoshop soft proof will mask areas that are out of gamut but has significant error tolerance and requires the colors be over about 6dE.

BTW, black is also out of gamut for all printers since no ink can adsorb 100% of light.