Pages: [1] 2 3 ... 5   Go Down

Author Topic: Some questions about working with Absolute Colorimetric intents  (Read 16215 times)

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

I'm sure this is probably a Photoshop Color Management 101 question, but since my last basic question resulted in me discovering that my Canon glossy paper profiles were broken, I figured it can't hurt to make sure I properly understand a few things regarding Absolute Colorimetric rendering intents.

Quick context for the questions: My goal has been to duplicate the appearance of some old photos as close to the original as possible. I think I made the mistake of asking a too-detailed question about color casts in my scans instead of breaking it down into core components first and checking my understanding of each phase.  My follow-up question about the broken rendering intents of my Canon paper profiles helped me realize that printing my scans using a functioning Absolute rendering intent improved my results by emulating the paper white of the source photos onto the paper that I have at my disposal.


As a test, I'm currently working with a simulated ColorChecker image from BabelColor, since its Lab values match the reference file of the physical ColorChecker Mini that I have and I can tell what they're supposed to be.


In Photoshop under Edit > Color Settings

Does this just control how the current image profile is translated into my display profile and what the Info panel numbers represent? The default is Relative, but when doing reproduction-style work, is it best to leave this in Absolute Colorimetric to make it easier to check that profiles are working properly?


Using Edit > Convert to Profile vs Soft Proofing

If I take this ColorChecker image and convert from Lab to sRGB using Absolute Colorimetric for example, most of the patches stay the same since they're almost all within the sRGB gamut, the cyan shifts a bit.  What I'm confused by is that if I convert from Lab to most of my paper profiles using Absolute Colorimetric, is that the patches will obviously shift. For example: If I choose a semi-gloss paper, the L of the white patch goes down, but the RGB values go up and the image gets brighter on the screen as a result. It seems counter-inuitive. Does this have something to do with how it preserves the white point of the source relative to the destination profile? (I can include photos if my description is gibberish.)

If I take the ColorChecker and proof it using Absolute Intent, unlike Converting the Lab values all stay the same but the image on the screen will clearly change depending on my settings for simulating paper color, etc. Is the proof just trying to emulate what the actual Lab values will look like when printed vs. the Convert reflecting the actual changes that must take place to preserve the appearance?  I just want to be clear on the distinction going forward.


Finally… evaluating the print

If I print the ColorChecker image using Absolute Colorimetric, I think I read in a few posts that it's possible to sample the output using a spectrophotometer and compare to the original. Is that correct? If so, what should the numbers be a match to – the reference Lab values, or the values after converting to the paper profile using an Absolute intent? Or can I really only just view it by eye to evaluate?

The sample I printed yesterday on my Pro-10 (from Windows with a working paper profile) using Absolute Colorimetric looks nice and vibrant with a slightly off-white white patch and the greys appear pretty neutral, though my glossy print vs the physical ColorChecker obviously reflect light differently.

For the heck of it, I did use spotread from ArgyllCMS to measure my printed patches…

Printed target using Absolute Colorimetric:
Canon glossy paper white:  94.055, -0.611, -2.047
ColorChecker white:  92.887, -1.425, 3.816
ColorChecker grey (under the yellow):  47.370, -0.039, -0.405
ColorChecker black:  16.089, 0.091, -1.236

Target Reference Lab values:
White:  95.19, -1.030, 2.930
Grey:  50.76, -0.116, 0.141
Black:  20.64, 0.070, -0.460

Target converted to the glossy paper profile using Absolute Colorimetric:
White:  93.31, -1.147, 2.273
Grey:  51.04, 0.164, -0.444
Black:  20.45, -0.077, -1.827




Thanks for any feedback!




*** 5/12/2018 - Edited the title
« Last Edit: May 12, 2018, 04:52:52 pm by BrianToth »
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/

Let’s start with the color settings and move on:

Photoshop CC's Color Settings & the Convert to Profile and Assign Profile command.

This new video covers everything you thought you wanted to know about the Photoshop Color Setting dialog. It also discusses the Convert to Profile Command and the Assign Profile Command. Photoshop CC 2017 is used in this video and it updates the video on this subject I Published on June 28, 2012.

High Rez: http://digitaldog.net/files/PhotoshopColorSettings.mp4
Low Rez (YouTube): https://www.youtube.com/watch?v=6JaHOGDK5OI
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

I'm sure this is probably a Photoshop Color Management 101 question, but since my last basic question resulted in me discovering that my Canon glossy paper profiles were broken, I figured it can't hurt to make sure I properly understand a few things regarding Absolute Colorimetric rendering intents.

Quick context for the questions: My goal has been to duplicate the appearance of some old photos as close to the original as possible. I think I made the mistake of asking a too-detailed question about color casts in my scans instead of breaking it down into core components first and checking my understanding of each phase.  My follow-up question about the broken rendering intents of my Canon paper profiles helped me realize that printing my scans using a functioning Absolute rendering intent improved my results by emulating the paper white of the source photos onto the paper that I have at my disposal.
This is exactly what Abs. Col. Intent was designed for.
Quote


As a test, I'm currently working with a simulated ColorChecker image from BabelColor, since its Lab values match the reference file of the physical ColorChecker Mini that I have and I can tell what they're supposed to be.


In Photoshop under Edit > Color Settings

Does this just control how the current image profile is translated into my display profile and what the Info panel numbers represent? The default is Relative, but when doing reproduction-style work, is it best to leave this in Absolute Colorimetric to make it easier to check that profiles are working properly?
This should just be set on Relative Colorimetric. It doesn't affect printing or soft proofing.
Quote



Using Edit > Convert to Profile vs Soft Proofing

If I take this ColorChecker image and convert from Lab to sRGB using Absolute Colorimetric for example, most of the patches stay the same since they're almost all within the sRGB gamut, the cyan shifts a bit.  What I'm confused by is that if I convert from Lab to most of my paper profiles using Absolute Colorimetric, is that the patches will obviously shift. For example: If I choose a semi-gloss paper, the L of the white patch goes down, but the RGB values go up and the image gets brighter on the screen as a result. It seems counter-inuitive. Does this have something to do with how it preserves the white point of the source relative to the destination profile? (I can include photos if my description is gibberish.)
Don't convert to paper profiles. There are very few reasons to do so unless needed to send a converted image file to an outside printer. Also, the way Photoshop displays images converted to a printer profile is messed up, To see what the printer prints, use the soft proof dialog.
Quote


If I take the ColorChecker and proof it using Absolute Intent, unlike Converting the Lab values all stay the same but the image on the screen will clearly change depending on my settings for simulating paper color, etc. Is the proof just trying to emulate what the actual Lab values will look like when printed vs. the Convert reflecting the actual changes that must take place to preserve the appearance?  I just want to be clear on the distinction going forward.
If you soft proof using Abs. Col., select the show paper color. Otherwise the image will brighten overall. This is because without the paper color checked, it uses Rel. Col. in the reversal process and will boost brightness. The lower the paper's white point, the more the boost.
Quote



Finally… evaluating the print

If I print the ColorChecker image using Absolute Colorimetric, I think I read in a few posts that it's possible to sample the output using a spectrophotometer and compare to the original. Is that correct? If so, what should the numbers be a match to – the reference Lab values, or the values after converting to the paper profile using an Absolute intent? Or can I really only just view it by eye to evaluate?
After converting to a paper profile, in order to correctly view the colors printed, make a copy and convert it back to Lab us Abs. Col. This avoids any interaction with the color settings menu which can result in incorrect values when displaying an image in the printer profile colorspace.  The values shown after converting represent the profile's A2B tables estimate of the color printed. If they are off more than 1 dE then it's a safe bet that color is out of the printer's gamut.

A spectro is the best way to test it.
Quote



The sample I printed yesterday on my Pro-10 (from Windows with a working paper profile) using Absolute Colorimetric looks nice and vibrant with a slightly off-white white patch and the greys appear pretty neutral, though my glossy print vs the physical ColorChecker obviously reflect light differently.

For the heck of it, I did use spotread from ArgyllCMS to measure my printed patches…

Printed target using Absolute Colorimetric:
Canon glossy paper white:  94.055, -0.611, -2.047
ColorChecker white:  92.887, -1.425, 3.816
ColorChecker grey (under the yellow):  47.370, -0.039, -0.405
ColorChecker black:  16.089, 0.091, -1.236
These are very low L* numbers for a CC. You might want to have your spectro calibrated. The a* and b* (neutrality) are within reason.
Quote


Target Reference Lab values:
White:  95.19, -1.030, 2.930
Grey:  50.76, -0.116, 0.141
Black:  20.64, 0.070, -0.460

Target converted to the glossy paper profile using Absolute Colorimetric:
White:  93.31, -1.147, 2.273
Grey:  51.04, 0.164, -0.444
Black:  20.45, -0.077, -1.827
These are consistent aside from the anomaly of the instrument's low L* reading of CC white patches.
« Last Edit: April 17, 2018, 11:09:37 pm by Doug Gray »
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/

If you don't know what or why you'd use Absolute Colorimetric rendering intent, don't use it  ;D .
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

A spectro is the best way to test it.These are very low L* numbers for a CC. You might want to have your spectro calibrated. The a* and b* (neutrality) are within reason.These are consistent aside from the anomaly of the instrument's low L* reading of CC white patches.

Thanks Doug.  Do you suppose it's possible my readings of my prints could be affected by brighteners in the Canon Glossy II paper I'm using? I have a brand new i1Studio (formerly ColorMunki) spectrophotometer, which is obviously an "entry level" piece of equipment, but I would hope it to be within certain specs and not too wildly off. I did recently read though that the i1Studio has a UV filter and might be useless with papers that use OBAs. I don't have another spectrophotometer to test against, and all I have to try to sample with "known" reference values are my targets.  If I attempt a spot-check on my physical ColorChecker mini with the matte patches…

Argyll Spotread     vs.     X-Rite Reference File
White:  96.269, -0.134, 1.139     vs.     95.19, -1.030, 2.930     =  CIE 2000: 2.14
Grey:  50.718, -0.296, -0.331     vs.     50.76, -0.116, 0.141     =  CIE 2000: 0.54
Black:  20.586, 0.343, -0.320     vs.     20.64, 0.070, -0.460     = CIE 2000: 0.43

Though this isn't exactly a perfect test either as there's a margin of error between X-Rite's target manufacturing, and their reading, and my readings even if everything were working. I would really need an individually measured target and reference I would think. But the results are closer than my readings from the paper.
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

Forgot to do the obvious test, and just measure my reading of the paper vs the profile's whitepoint:

My Argyll spotread on Canon Glossy II paper:  94.055, -0.611, -2.047
ICC profile's white point:  94.19024, -0.5772792, -2.4742979
100% Lab in Photoshop with relative conversion to paper profile:  94.19, -0.562, -2.491

So, not sure what to make of things now, looks like my readings aren't too bad?
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

Thanks Doug.  Do you suppose it's possible my readings of my prints could be affected by brighteners in the Canon Glossy II paper I'm using? I have a brand new i1Studio (formerly ColorMunki) spectrophotometer, which is obviously an "entry level" piece of equipment, but I would hope it to be within certain specs and not too wildly off. I did recently read though that the i1Studio has a UV filter and might be useless with papers that use OBAs. I don't have another spectrophotometer to test against, and all I have to try to sample with "known" reference values are my targets.  If I attempt a spot-check on my physical ColorChecker mini with the matte patches…

Argyll Spotread     vs.     X-Rite Reference File
White:  96.269, -0.134, 1.139     vs.     95.19, -1.030, 2.930     =  CIE 2000: 2.14
Grey:  50.718, -0.296, -0.331     vs.     50.76, -0.116, 0.141     =  CIE 2000: 0.54
Black:  20.586, 0.343, -0.320     vs.     20.64, 0.070, -0.460     = CIE 2000: 0.43

Though this isn't exactly a perfect test either as there's a margin of error between X-Rite's target manufacturing, and their reading, and my readings even if everything were working. I would really need an individually measured target and reference I would think. But the results are closer than my readings from the paper.

The i1studio doesn't use a uV cut filter, rather it uses a white led that has virtually no uV in it. It's the same technology that the ColorMunki and iSiS uses for generating M2 profiles.

IMO, M2 profiles are better for most uses than M0 or M1 profiles these days for indoor use because most lighting no longer uses incandescent lamps and have lower or no uV. M1 is really best for displaying prints outdoors and M0 for natural lighting though a window.

Because of this shift in lighting tech., indoor lighting will, in most cases, be a better match for M2 profiles regardless of whether a paper has OBAs or not. And for papers with no OBAs there is no difference between M0, M1, or M2 profiles. They should produce identical results.

I think your spectro is working correctly. The lower b* reading on the CC white patch is not uncommon. Most CCs have a b* between 2 and 3 but I've seen numbers around 1 as well.  The other neutral patches are quite reasonable and are also more consistent between different CCs. So that's a good sign things are working spectro wise.

The problem may be in your profile or printer driver settings. You should expect lower dE numbers from the CC image print using Abs. Col. How did you make or acquire the profile you are using?
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

Forgot to do the obvious test, and just measure my reading of the paper vs the profile's whitepoint:

My Argyll spotread on Canon Glossy II paper:  94.055, -0.611, -2.047
ICC profile's white point:  94.19024, -0.5772792, -2.4742979
100% Lab in Photoshop with relative conversion to paper profile:  94.19, -0.562, -2.491

So, not sure what to make of things now, looks like my readings aren't too bad?
Perfectly good numbers. Very typical of the small variations seen on glossy paper. Also, the slightly negative b* readings are also typical of most papers. It's from a colorant manufacturers add to make the paper look less warm. BTW, printing L*=100 with Colorimetric intent (or anything other than Abs.) should result in nothing printing at all so those numbers are all good.
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

The problem may be in your profile or printer driver settings. You should expect lower dE numbers from the CC image print using Abs. Col. How did you make or acquire the profile you are using?

My methodology has been:
  • Open the attached BabelColor Lab ColorChecker TIFF in Photoshop CC 2018.
  • Print with Photoshop managed colors in Absolute, choosing the Canon-supplied Photo Paper Plus Glossy II profile.
  • Under the Canon Pro-10 printer settings in Windows I made sure to go to Manual Color Adjustment > Matching and set it to None. I then choose the correct paper type, and set the quality to high, leaving all other settings alone. (On the Mac when I print I can just choose Photoshop managed color and it'll disable the printer driver settings automatically, but the Mac version of Photoshop isn't working with those particular profiles anyway.)
  • Print and wait until the ink is dry [enough].
  • Measure using Argyll's spotread tool and the i1Studio, doing multiple readings to eliminate outliers.

I'm currently doing some similar tests with some different papers that I have profiles for and that work on the Mac too and I'll see what I get for results.


Perfectly good numbers. Very typical of the small variations seen on glossy paper. Also, the slightly negative b* readings are also typical of most papers. It's from a colorant manufacturers add to make the paper look less warm. BTW, printing L*=100 with Colorimetric intent (or anything other than Abs.) should result in nothing printing at all so those numbers are all good.

That particular non-absolute conversion test was just a Photoshop sanity check, but I didn't actually try printing it. Maybe I should have. (Glad to know I'm not crazy though, just a bit ignorant when it comes to Photoshop and color science. This project has been a crash course in both, does help a bit having a computer science background.)
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

Canon Pro Platinum Glossy

Used Canon supplied ICC profile for Canon Pro-10.  All measurements were done against a background of black construction paper. (Didn't notice a notable difference against a white background though.)

ICC profile white point:     96.2402999, -0.6663628, 0.9857583
Measured paper white:     95.328430 -0.796546 1.600733     CIE2000: 0.81     CIE1979: 1.11


ColorChecker Lab values after converting to paper profile using absolute colorimetric in Photoshop:
White:  95.11  -1.022  2.930
Grey:  51.23  0.172  -0.155
Black:  20.43  -0.491  -1.983


Printed using absolute 16-bit from Photoshop CC 2018 on Mac:
White:  93.909180  -1.209533  3.318551     CIE2000: 0.83     CIE1976: 1.28
Grey:  46.562440  -0.001105  -0.874291     CIE2000: 4.71     CIE1976: 4.73
Black:  15.956659  0.485438  -1.726592     CIE2000: 3.37     CIE1976: 4.59


Printed using absolute 8-bit from Photoshop CC 2018 on Windows 10:   
White:  93.836831  -1.229150  3.539094     CIE2000: 0.97     CIE1976: 1.43
Grey:  47.579424  0.216308  -0.602152     CIE2000: 3.76     CIE1976: 3.68
Black:  15.992631  0.530592  -1.728224     CIE2000: 3.38     CIE1976: 4.56


Printed using absolute 8-bit from Photoshop CC 2018 on Windows 10 and measured a week (or 2) later:   
White:  93.785744  -1.421952  3.711370     CIE2000: 1.16     CIE1976: 1.59
Grey:  48.070289  0.230349  -0.587270     CIE2000: 3.24     CIE1976: 3.21
Black:  16.604610  0.528541  -1.519456     CIE2000: 3.04     CIE1976: 3.9

Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

Finally, Red River Pecos River Gloss

Used ICC profile from Red River for Canon Pro-10, following their instructions for Canon paper/quality driver settings.  All measurements were done against a background of black construction paper. (Didn't notice a notable difference against a white background though.)

ICC profile white point:     96.3303456, 0.3228647, -0.6502543
Measured paper white:     95.170087 0.094665 -1.445915     CIE2000:  1.09     CIE1976:  1.43

ColorChecker Lab values after converting to paper profile using absolute colorimetric in Photoshop:
White:     94.69  -0.937  2.727
Grey:     50.76  -0.093  0.164
Black:     20.69  0.156  -0.499

Printed using absolute 16-bit from Photoshop CC 2018 on Mac (measured later that day):
White:     92.771324 -1.153834 1.552349     CIE2000:  1.62     CIE1976:  2.26
Grey:     48.815054 -0.191202 -1.126463     CIE2000:  2.32     CIE1976:  2.34
Black:     18.736083 0.498224 -0.592850     CIE2000:  1.44     CIE1976:  1.99


Printed using absolute 8-bit from Photoshop CS6 on Mac (measured later that day):   
White:     92.857677 -0.944758 1.192230     CIE2000:  1.8     CIE1976:  2.39
Grey:     47.703535 -0.106346 -1.433525     CIE2000:  3.42     CIE1976:  3.45
Black:     18.039598 0.571037 -0.802760     CIE2000:  1.94     CIE1976:  2.7


Printed using absolute 8-bit from Photoshop CS6 on Mac (measured next day):   
White:     92.696839 -1.203567 1.527344     CIE2000:  1.69     CIE1976:  2.34
Grey:     48.613465 -0.123885 -1.223072     CIE2000:  2.54     CIE1976:  2.56
Black:     18.690034 0.484400 -0.521634     CIE2000:  1.46     CIE1976:  2.03


Forgot to print from Windows, probably wouldn't deviate too much… wanted to compare different versions of Photoshop.
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

I general yes, avoid it. But there are some appropriate places it is needed.

This explains where Abs. can be of use when reproducing scanned images:
https://books.google.com/books?id=OxlBqY67rl0C&pg=PA290&lpg=PA290&dq=scanners+%22absolute+colorimetric%22+relative&source=bl&ots=IyXihwQSog&sig=mcJZcCFx9P1uCCjSfyvK_Sv2qC8&hl=en&sa=X&ved=2ahUKEwi68N65yO_aAhUM4IMKHZpDD_U4FBDoATAAegQIABAn#v=onepage&q=scanners%20%22absolute%20colorimetric%22%20relative&f=false

Thank you for the link to that section of that book. I found a couple other good books on the subject, but everyone words things a bit differently and I thought that one you shared gave a good example of how the two work.

I was initially skeptical of using Absolute at all because most intro discussions to the topic of rendering intents outright say "don't use it" or they ignore it altogether. Lightroom doesn't even give an option to use it, which is probably reasonable when printing digital photos that aren't trying to reproduce existing artwork. (Clearly in most cases you'd probably want the white of the paper to match the white of what's being printed.) And if I print using any other apps, the printer drivers don't give any options at all for intents, so I'm not sure what method the driver uses internally if you set it to use a profile. (But that doesn't matter right now.) Photoshop just says "suitable for proofing to simulate the output of a particular device" which to me wasn't clear that it would be useful for preserving the actual colors of my scanned images.

So, I feel I'm understanding the subject. But when my input and my output results don't match what my understanding is, I start to second guess all the basics. Which is why I started this thread. :)
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

Finally, Red River Pecos River Gloss

Used ICC profile from Red River for Canon Pro-10, following their instructions for Canon paper/quality driver settings.  All measurements were done against a background of black construction paper. (Didn't notice a notable difference against a white background though.)

ICC profile white point:     96.3303456, 0.3228647, -0.6502543
Measured paper white:     95.170087 0.094665 -1.445915     CIE2000:  1.09     CIE1976:  1.43

ColorChecker Lab values after converting to paper profile using absolute colorimetric in Photoshop:
White:     94.69  -0.937  2.727
Grey:     50.76  -0.093  0.164
Black:     20.69  0.156  -0.499

Printed using absolute 16-bit from Photoshop CC 2018 on Mac (measured later that day):
White:     92.771324 -1.153834 1.552349     CIE2000:  1.62     CIE1976:  2.26
Grey:     48.815054 -0.191202 -1.126463     CIE2000:  2.32     CIE1976:  2.34
Black:     18.736083 0.498224 -0.592850     CIE2000:  1.44     CIE1976:  1.99


Printed using absolute 8-bit from Photoshop CS6 on Mac (measured later that day):   
White:     92.857677 -0.944758 1.192230     CIE2000:  1.8     CIE1976:  2.39
Grey:     47.703535 -0.106346 -1.433525     CIE2000:  3.42     CIE1976:  3.45
Black:     18.039598 0.571037 -0.802760     CIE2000:  1.94     CIE1976:  2.7


Printed using absolute 8-bit from Photoshop CS6 on Mac (measured next day):   
White:     92.696839 -1.203567 1.527344     CIE2000:  1.69     CIE1976:  2.34
Grey:     48.613465 -0.123885 -1.223072     CIE2000:  2.54     CIE1976:  2.56
Black:     18.690034 0.484400 -0.521634     CIE2000:  1.46     CIE1976:  2.03


Forgot to print from Windows, probably wouldn't deviate too much… wanted to compare different versions of Photoshop.

These numbers just indicate that the profiles you are using don't quite match your printer's response. Consumer printers tend to vary a bit, perhaps as much as 3 dE or more from each other. The higher end printers are individually tuned to match a reference response but this is less practical for the high volume consumer or semi-pro printers. And printers age and change over time. Normally, people don't notice these differences printing regular photos but it becomes more apparent when people are trying to match colors. And that, of course, is why Abs. Col. exists. This is one of the reasons people buy printer profiling hardware.

You should be profiling the papers you use, not using profiles from the paper vendors or printer OEMs. You will then get much better results.

BTW, when you create a profile you should be able to use the same one for Windows as well as iOS and it shouldn't make any difference what computer you make them on. Use your XRite software, it's quite good.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

Finally, Red River Pecos River Gloss

Used ICC profile from Red River for Canon Pro-10, following their instructions for Canon paper/quality driver settings.  All measurements were done against a background of black construction paper. (Didn't notice a notable difference against a white background though.)

ICC profile white point:     96.3303456, 0.3228647, -0.6502543
Measured paper white:     95.170087 0.094665 -1.445915     CIE2000:  1.09     CIE1976:  1.43

ColorChecker Lab values after converting to paper profile using absolute colorimetric in Photoshop:
White:     94.69  -0.937  2.727
Grey:     50.76  -0.093  0.164
Black:     20.69  0.156  -0.499

Exactly what process are you doing by converting Abs. Col. in Photoshop and what are you expecting from that process? If you are going through the transform of a working RGB space to a printer RGB space then back to a working space the transform will produce small changes based on rounding and interpolation errors in the ICC profiles for all colors that are in gamut (printable)*.  This tells you nothing about how accurate the profile is, only that it is well formed. The numbers will typically change very little.

The only way to test a profile's accuracy is print known colors and measure them. If you print using Abs. Col. and a custom profile then measure them you should have average dEs under 1.0.

Also, ICC profiles are based on measurements using a white backing. This differs from some parts of graphic arts where a black backing is used. for best match when measuring print colors, use a white backing.

* If you, in Photoshop, take an image in a working space and convert it to the printer space via "Convert Profile" you can select whatever ICC intent you wish including BPC. Once you have that and wish to know what colors the printer profile expects will be printed, you must convert the image back to Lab or a standard RGB working space that is at least as large as the printer's gamut. I recommend either Lab or ProPhoto RGB. When converting back, select Abs. Col. This resulting image will have the color values that it expects to be printed. How close that is depends on how good your profile is and generally custom profiles produce the best match.

BTW, this a a good way to see what gets printed when you start with colors outside a printer's gamut. It will tell you quite closely what the actual, printable color is expected to be.
« Last Edit: May 06, 2018, 08:05:06 pm by Doug Gray »
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

Hey Doug, thanks for taking the time and having the patience to provide all the details you have. I really appreciate it. I did a little experiment based on what you shared with me, just to make sure I'm understanding completely:

Given I print an image with known color values (in this case the "synthetic" ColorChecker from BabelColor) using a particular rendering intent…

If I want to reliably evaluate the accuracy of the print, I can:
  • Take the image I printed and convert it to the same printer paper profile using the same rendering intent settings.
  • Convert the image to Lab using absolute colorimetric.
  • Measure the print with my spectrophotometer and compare the Lab values to the Lab values that I see in Photoshop's info panel. (In the attached image those would be the values highlighted in green or blue.)

Assuming I'm on the same page as you :) then at least I know that my comparisons above were done correctly. I have my Photoshop color settings currently set to absolute colorimetric and the image I've been working with is natively in the Lab color space and I've only been testing absolute printing, so hopefully I haven't muddied anything up.
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43

These numbers just indicate that the profiles you are using don't quite match your printer's response.



You should be profiling the papers you use, not using profiles from the paper vendors or printer OEMs. You will then get much better results.

I had expected a certain degree of error with OEM profiles, given manufacturing tolerances, etc., but never quite expected it to be that off. Using the OEM profiles probably wouldn't really be a concern for regular photo printing for an amateur like me, the Pro-10 is already much better than most photo printers I've used over the years. But I did as you suggested yesterday and waited until today to measure my results…

The OEM Canon Glossy II profile gave me an average ∆E (1976) of 3.681.

My custom i1Studio profile brought that down to an average ∆E of 1.312!  It's possible I could get that closer with more patience. – For example I only waited about 30 minutes for dry time for my profiling, perhaps I could've let that settle down a bit more. It'd be nice if i1Studio gave me some data to look at. If I use Argyll I can double-check my readings before making the profile… I may try that down the road.

An interesting observation:  The printed ColorChecker using the OEM Canon profile had much more saturated colors, almost like it was setup to give the output more punch. The colors are toned down using the custom profile.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

Assuming I'm on the same page as you :) then at least I know that my comparisons above were done correctly. I have my Photoshop color settings currently set to absolute colorimetric and the image I've been working with is natively in the Lab color space and I've only been testing absolute printing, so hopefully I haven't muddied anything up.

Your Photoshop methodology is dead on. Those numbers are exactly what I would expect. Even the difference when using sRGB tiffs compared to Lab. With sRGB you are hitting up against the sRGB gamut edge and that accounts for the different L* drops for RI. as well as the saturation decrease. Nice work.

if you use ProPhoto RGB space you won't see the differences you get with sRGB.
« Last Edit: May 09, 2018, 12:30:11 am by Doug Gray »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

I had expected a certain degree of error with OEM profiles, given manufacturing tolerances, etc., but never quite expected it to be that off. Using the OEM profiles probably wouldn't really be a concern for regular photo printing for an amateur like me, the Pro-10 is already much better than most photo printers I've used over the years. But I did as you suggested yesterday and waited until today to measure my results…

The OEM Canon Glossy II profile gave me an average ∆E (1976) of 3.681.

My custom i1Studio profile brought that down to an average ∆E of 1.312!  It's possible I could get that closer with more patience. – For example I only waited about 30 minutes for dry time for my profiling, perhaps I could've let that settle down a bit more. It'd be nice if i1Studio gave me some data to look at. If I use Argyll I can double-check my readings before making the profile… I may try that down the road.

An interesting observation:  The printed ColorChecker using the OEM Canon profile had much more saturated colors, almost like it was setup to give the output more punch. The colors are toned down using the custom profile.

The high dE numbers for the canned profiles is pretty typical. So is the higher saturation, My Canon also produces higher saturation using canned profiles. My suspicion is that Canon is trying to create more "Pop."  It's even worse on my Canon if you let the printer manage color.

As for an average dE76 of 1.3 for your i1Studio profiled print, that's not bad really. Also, you will likely find that if you use dE2000 the average is quite a bit lower. Likely around .8.  i1Studio has, I believe, an option for a second pass where they print colors based on the first set that improve the profile. This should improve things further.

Also, don't worry too much about dry time. Most of the changes occur in the first 10 minutes of drying. After 30 minutes dE errors from print drying become a fraction of the total and likely accounts for no more than 1/3 of the dEs.  After 2 hours around 1/5th of the dEs compared to a two day dry time. Of course it's best to let prints dry 24 hrs to make the best profiles but don't expect significant differences.

I posted an extensive series of measurements on drying time color shifts some months back.

BTW, if you want to get a sense of how gamut mapping works at and beyond the printer's gamut boundary, try this:

In Photoshop create a gradient from Lab (50,-120,0) to (50,0,0). Then convert it using Abs. Col. to Printer space and back again to Lab. Now run the cursor down the gradient and watch the small variations of b* around 0 until you get to the place where it can no longer print the requested colors. What it actually prints will be reasonably accurately reported and you will see shifts in the Lab values as it maps the OOG colors to the nearest point on the 3D gamut space.  It can be interesting to do the same thing but converting to Perceptual, then back to Lab using Abs. You will see the different ways Perc. maps colors.
« Last Edit: May 09, 2018, 01:22:18 am by Doug Gray »
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4344
    • A Goldhammer Photography


My custom i1Studio profile brought that down to an average ∆E of 1.312!  It's possible I could get that closer with more patience. – For example I only waited about 30 minutes for dry time for my profiling, perhaps I could've let that settle down a bit more. It'd be nice if i1Studio gave me some data to look at. If I use Argyll I can double-check my readings before making the profile… I may try that down the road.
Just as a reference point, I use Argyll to do all my printer profiling (Epson 3880) and my deltaE is usually right around 1.0.  Profiles are quite smooth when visualized.
Logged
Pages: [1] 2 3 ... 5   Go Up