Pages: 1 [2] 3   Go Down

Author Topic: Choices Between Printer Color Profiling Engines  (Read 4226 times)

andrewrodney

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13226
    • http://www.digitaldog.net/
Re: Choices Between Printer Color Profiling Engines
« Reply #20 on: December 18, 2017, 09:31:57 PM »


Interesting. So you are suggesting bypassing the PCS and going directly from, say Adobe RGB or one of the larger spaces that encompasses printable colors? This would lock in a working space but yeah, it should work better than going through the PCS. I think of device link profiles as gamut mapping from one output device to another output device but yeah, no reason it couldn't go from RGB tri color working spaces to a printer colorimetrically.
That's where Device Links profiles come in so handy; no PCS necessary/conversions can be direct such as CMYK to CMYK. A device link is like two profiles hardwired into one. There is no Source and Destination methodology with device links because they are built as a one-way street, allowing only a very specific, fixed type of conversion.
Logged
Andrew Rodney
Author ďColor Management for Photographers"

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1154
Re: Choices Between Printer Color Profiling Engines
« Reply #21 on: December 18, 2017, 10:22:16 PM »

That's where Device Links profiles come in so handy; no PCS necessary/conversions can be direct such as CMYK to CMYK. A device link is like two profiles hardwired into one. There is no Source and Destination methodology with device links because they are built as a one-way street, allowing only a very specific, fixed type of conversion.
True! And I always just thought of device link profiles as something the CYMK peeps did for their advantage in controlling inking. I just never thought of using device link profiles to go from, say sRGB or Adobe RGB to the printer directly. But there are obvious advantages. Especially for sRGB photos with perceptual mapping where you know the source profile was sRGB and can map into the printer's profile to create more attractive prints than are possible when using standard profiles. Also, all but the largest RGB colorspaces should produce significantly less error in colorimetric conversions. OTOH, going from a space like ProPhoto would increase the distance between LUT grid points. It isn't clear to me that a ProPhoto RGB link to a printer would be better than going through the standard ppRGB->PCSLAB->printer. Would be interesting to create and test device link profiles against standard profiles when using wide RGB working spaces.

EtoA: I just tried making a device link profile with I1Profiler RGB->RGB and it only allows source profiles that are printer type devices.
« Last Edit: December 18, 2017, 10:28:43 PM by Doug Gray »
Logged

Ethan Hansen

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 92
    • Dry Creek Photo
Re: Choices Between Printer Color Profiling Engines
« Reply #22 on: December 18, 2017, 10:42:21 PM »

True! And I always just thought of device link profiles as something the CYMK peeps did for their advantage in controlling inking. I just never thought of using device link profiles to go from, say sRGB or Adobe RGB to the printer directly. But there are obvious advantages. Especially for sRGB photos with perceptual mapping where you know the source profile was sRGB and can map into the printer's profile to create more attractive prints than are possible when using standard profiles. Also, all but the largest RGB colorspaces should produce significantly less error in colorimetric conversions. OTOH, going from a space like ProPhoto would increase the distance between LUT grid points. It isn't clear to me that a ProPhoto RGB link to a printer would be better than going through the standard ppRGB->PCSLAB->printer. Would be interesting to create and test device link profiles against standard profiles when using wide RGB working spaces.

EtoA: I just tried making a device link profile with I1Profiler RGB->RGB and it only allows source profiles that are printer type devices.

i1Profiler's devicelink implementation isn't particularly useful. THere isn't much control and, as you discovered, there is no way to link dissimilar profile types.

If your source image is in a huge color space such as PP RGB, the way to make best use of device links is via Argyll's ability to create bespoke, image-specific DL's. This is something we do with regularity when working with the wonky printers and substrates used for building wraps or architectural mesh. A 30 story tall image can easily highlight transition lines or banding. Making a devicelink based on the actual color values in the source image helps considerably.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1154
Re: Choices Between Printer Color Profiling Engines
« Reply #23 on: December 18, 2017, 11:05:06 PM »

i1Profiler's devicelink implementation isn't particularly useful. THere isn't much control and, as you discovered, there is no way to link dissimilar profile types.

If your source image is in a huge color space such as PP RGB, the way to make best use of device links is via Argyll's ability to create bespoke, image-specific DL's. This is something we do with regularity when working with the wonky printers and substrates used for building wraps or architectural mesh. A 30 story tall image can easily highlight transition lines or banding. Making a devicelink based on the actual color values in the source image helps considerably.

I don't have a need like that. Generally I want to convert colorimetrically from some arbitrary RGB space, often ppRGB, to printer RGB. I would like to do it accurately.

I'm coming to the conclusion the simplest way, at least for me, would be to use the A2B tables and do a steepest descent search for the RGB values that produces the desired LAB. Alternately, just read in the RAW target RGB/LAB values and do some 3D interpolation in Matlab.

One thing I recently discovered is that the printer's repeatability is considerably higher with matte papers than it is with glossy types. Not sure why that is but it's been highly repeatable in experiments I've been doing the last week with similar results from both printers. So reproducing colors on matte paper would likely benefit from using a more refined technique that just depending on the B2A tables.
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: Choices Between Printer Color Profiling Engines
« Reply #24 on: December 18, 2017, 11:08:53 PM »

So you are suggesting bypassing the PCS and going directly from, say Adobe RGB or one of the larger spaces that encompasses printable colors?
Some sort of PCS is always going to be used in the linking process when the input to the process is ICC device profiles. But there is no necessity to use the B2A table.
« Last Edit: December 18, 2017, 11:19:26 PM by GWGill »
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: Choices Between Printer Color Profiling Engines
« Reply #25 on: December 18, 2017, 11:19:04 PM »

I'm coming to the conclusion the simplest way, at least for me, would be to use the A2B tables and do a steepest descent search for the RGB values that produces the desired LAB.
One of many possible approaches if it's an RGB profile. A numerical non-linear equation solver would be another.
More complicated if it's CMYK and one wishes to discover the whole locus of possible solutions though :-)
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1154
Re: Choices Between Printer Color Profiling Engines
« Reply #26 on: December 18, 2017, 11:36:25 PM »

One of many possible approaches if it's an RGB profile. A numerical non-linear equation solver would be another.
More complicated if it's CMYK and one wishes to discover the whole locus of possible solutions though :-)
Yes. glad I'm using RGB. Finding the full set of 4+ variables that map to a 3D LAB value sounds painful. Though, for people minimizing ink use or improving some other metric perhaps related to the screen or diffusion pattern, it's of significant interest.
Logged

Brad Paulson

  • Full Member
  • ***
  • Offline Offline
  • Posts: 205
Re: Choices Between Printer Color Profiling Engines
« Reply #27 on: December 19, 2017, 01:14:18 AM »

Tracking along.  Thanks for all the links. Surprisingly I read or tried to read all those links before posting.

The one thing Iíd add is that ProPhoto 16 bit seems to be the image processing space of the day.  Iíve profiled two cameras from RAW to 16 bit ProPhoto 5 and 3 years ago.  I gave that up when I started converting RAW to 16 bit ProPhoto TIFFs in Phocus about 6 months ago.  I tried Lab in PS, was excited for a while, but concluded that on that platform anyway it didnít seem ready. Maybe I should go back to creating camera profiles and following it through start to finish. Maybe not  Part of this is aesthetics. 

For now, converting a very well done 16 bit ProPhoto TIFF to printer space, yanking it around in the meantime (see Tony Kuyperís luminousity curve PS plugins for a yanking image processing example), is at least what I am in this post trying to drill down on. 

I guess collink may not be the right way for me if it means processing images in my NEC monitor space (measured much better than Adobe RGB and advertised) and outputting through that space.  I had mistakenly thought earlier that was a guide to Argyll in rendering the original TIFF file to print, if that makes sense. 
« Last Edit: December 19, 2017, 03:42:33 AM by Brad Paulson »
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: Choices Between Printer Color Profiling Engines
« Reply #28 on: December 19, 2017, 06:40:46 AM »

I had mistakenly thought earlier that was a guide to Argyll in rendering the original TIFF file to print, if that makes sense.
None at all I'm afraid. Perhaps you should be more explicit as to the workflows you are referring to.
Logged

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 743
Re: Choices Between Printer Color Profiling Engines
« Reply #29 on: December 19, 2017, 09:50:28 AM »

I always just thought of device link profiles as something the CYMK peeps did for their advantage in controlling inking.
Film "looks" and simulations we use in RPP and elsewhere are "device links". Some are RGB-RGB, some are Lab-Lab. Turned out, it is a superior method compared to just film "profiles", gets closer to the desired look, with negligible artifacts and very predictable behavior.
Logged

andrewrodney

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13226
    • http://www.digitaldog.net/
Re: Choices Between Printer Color Profiling Engines
« Reply #30 on: December 19, 2017, 10:31:44 AM »

The one thing Iíd add is that ProPhoto 16 bit seems to be the image processing space of the day.
There's another potential  benefit to very large RGB working space color gamuts like ProPhoto RGB when your goal is printing.

Simple matrix profiles of RGB working spaces when plotted 3 dimensionally illustrate that they reach their maximum saturation at high luminance levels. The opposite is seen with print (output) color spaces. Printers produce color by adding ink or some colorant, while working space profiles are based on building more saturation by adding more light due to the differences in subtractive and additive color models. To counter this, you need a really big RGB working space like ProPhoto RGB again due to the simple size and to fit the round peg in the bigger square hole. RGB working spaces have shapes which are simple and predictable and differ greatly from output color spaces. Then there is the issue of very dark colors of intense saturation which do occur in nature and we can capture with many devices. Many of these colors fall outside Adobe RGB (1998) and when you encode into such a color space or smaller gamut, you clip the colors to the degree that smooth gradations become solid blobs in print, again due to the dissimilar shapes and differences in how the two spaces relate to luminance. So the advantage of ProPhoto isn't only about retaining all those out-of-gamut colors it's also about maintaining the dissimilarities between them, so that you can map them into a printable color space as gradations rather than ending up as blobs. 


Here is a link to a TIFF that I built to show the effect of the 'blobs' and lack of definition of dark but saturated colors using sRGB (Red dots) versus the same image in ProPhoto RGB (Green dots). The image was synthetic, a Granger Rainbow which contains a huge number of possible colors. You can see that the gamut of ProPhoto is larger as expected. But notice the clumping of the colored red vs. green dots in darker tones which are lower down in the plot. Both RGB working space were converted to a final output printer color space (Epson 3880 Luster).

http://www.digitaldog.net/files/sRGBvsPro3DPlot_Granger.tif
Logged
Andrew Rodney
Author ďColor Management for Photographers"

Brad Paulson

  • Full Member
  • ***
  • Offline Offline
  • Posts: 205
Re: Choices Between Printer Color Profiling Engines
« Reply #31 on: December 26, 2017, 08:21:08 PM »

Sorry for the delay, but in the interim I had to order and replaced a printhead.

Attached are the results of a 4257 patch test in Argyll, DropRGB, and Copra for comparison.  I have not tested i1P as it's now apparent I need to purchase the hardware.

Here's a list of some assorted thoughts and preliminary conclusions.  Feel free to comment yourself.

1. These results are particular to printing on a Z3200ps.  It has a somewhat smaller gamut (though more archival inks) than competing models from Canon or Epson, and the results on those machines would be interesting to see in comparison.  If you have another printer you probably should conduct your own tests.  Even maybe if you have another Z3200.  Probably different results apply across papers too.  Here I used Ilford Gold Fibre Gloss which has no OBAs, high Gamut and D-max

2. I used a 4537 patch chart that dried 3 days before scanning it and creating the icc profile.  It could have been improved by using a more perfect patch set described by Doug here.  I will follow that process hopefully next time.  But for visual purposes, I believe the consensus there was it doesn't matter.

3. Very few differences exist in the in gamut colors across all software platforms.  Significant differences, however, exist in how out of gamut colors print across the software platforms as well as between rendering intents.  If you're printing prints with significant out of gamut colors, the findings here appear at least to me to be significant.

4. Perceptual generally did better than Saturation, which did better then Relative Colorimetric.

5. Overall, on an equally weighted, blended basis, DropRGB appears to my eye to be the winner, followed closely by Argyll, then Copra.  If you weigh Perceptual much more heavily, however (as I do), then Argyll appears to me to be the winner.  DropRGB is extremely close and I had thought I liked that more at first, but I realized that Argyll performed more truly on the luminance scale in the Granger Rainbow test and that seems more important to me.  It's debatable I guess and I'd be interested in what others see in these images.

Absent some new and noteworthy i1P information, I think I'm going to use Argyll for the intermediate future anyway.  Not solely because of its Perceptual rendering, but it's customization and I'm actually excited to try out its image to print profile capabilities (discussed above).
« Last Edit: December 26, 2017, 09:12:30 PM by Brad Paulson »
Logged

Brad Paulson

  • Full Member
  • ***
  • Offline Offline
  • Posts: 205
Re: Choices Between Printer Color Profiling Engines
« Reply #32 on: December 26, 2017, 08:22:34 PM »

. . . and the Saturation pics.

I converted these to sRGB.  If anyone wants the ProPhoto or an Adobe RGB file, let me know.
« Last Edit: December 26, 2017, 08:26:43 PM by Brad Paulson »
Logged

aaronchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 520
Re: Choices Between Printer Color Profiling Engines
« Reply #33 on: December 26, 2017, 10:46:32 PM »

. . . and the Saturation pics.

I converted these to sRGB.  If anyone wants the ProPhoto or an Adobe RGB file, let me know.

Thanks for your great test.
If you want to, you can send me the measurement CGATS and I can generate an icc profile with i1P for you.

Thank you
Aaron Chan

Brad Paulson

  • Full Member
  • ***
  • Offline Offline
  • Posts: 205
Re: Choices Between Printer Color Profiling Engines
« Reply #34 on: December 26, 2017, 10:59:53 PM »

Thanks Aaron.  That would make it complete.  I'll PM you.

« Last Edit: December 26, 2017, 11:10:22 PM by Brad Paulson »
Logged

aaronchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 520
Re: Choices Between Printer Color Profiling Engines
« Reply #35 on: December 26, 2017, 11:24:23 PM »

Thanks Aaron.  That would make it complete.  I'll PM you.

please check pm
thx

Brad Paulson

  • Full Member
  • ***
  • Offline Offline
  • Posts: 205
Re: Choices Between Printer Color Profiling Engines
« Reply #36 on: December 27, 2017, 03:21:55 AM »

Thanks again Aaron.  I printed up the i1P samples and am going to let them dry overnight and shoot them tomorrow alongside the other samples, and will post the results.  For the moment though, Argyll remains the most impressive with the Z3200, followed by DropRGB.

Andrew, if youíre still around, Iím getting darker blue balls with i1P than I expected.  You might want to check that out in my pics tomorrow.  Not as dark/black as Copra, but much darker than those in your video, particularly in the darker ball.  I suspect itís more likely a result of the inkset used by the Z, or possibly that we used i1Pís default settings, than an error in color management as Iíve been over the latterís pipeline quite a few times.  But very interested in your take on that.  Iíll get a little better light on the final shot too because there is clear gradation in the i1P blue sample thatís not at all so clear in Copraís and the darks in all of them simply need more light in the shots. 
Logged

andrewrodney

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13226
    • http://www.digitaldog.net/
Re: Choices Between Printer Color Profiling Engines
« Reply #37 on: December 27, 2017, 01:46:04 PM »

Andrew, if youíre still around, Iím getting darker blue balls with i1P than I expected.  \
But not black right?
Logged
Andrew Rodney
Author ďColor Management for Photographers"

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1154
Re: Choices Between Printer Color Profiling Engines
« Reply #38 on: December 27, 2017, 02:34:31 PM »

Thanks again Aaron.  I printed up the i1P samples and am going to let them dry overnight and shoot them tomorrow alongside the other samples, and will post the results.  For the moment though, Argyll remains the most impressive with the Z3200, followed by DropRGB.

Andrew, if youíre still around, Iím getting darker blue balls with i1P than I expected.  You might want to check that out in my pics tomorrow.  Not as dark/black as Copra, but much darker than those in your video, particularly in the darker ball.  I suspect itís more likely a result of the inkset used by the Z, or possibly that we used i1Pís default settings, than an error in color management as Iíve been over the latterís pipeline quite a few times.  But very interested in your take on that.  Iíll get a little better light on the final shot too because there is clear gradation in the i1P blue sample thatís not at all so clear in Copraís and the darks in all of them simply need more light in the shots.

Hi Brad,

This is a topic worth further discussion as it shows some of the risks using a large colorspace that extends to imaginary colors and the "blue" ball on the bottom right of Andrew's Gamut_Test_File_Flat.tiff image represents the most extreme, and imaginary, RGB blue. Load the file in Photoshop, bring up the info panel, and select Lab as the colorspace to show. Move the cursor to different parts of the blue ball. Notice the L* stays below 1 (.1 actually - even for max blue) which is a luminance below all printable blacks. The ICCLAB value of ProPhotoRGB(0,0,255) is L*=.08, a*=90, b*=-128. (ICCLAB clips CIELAB at -128, Actual CIELAB b*=-172).

So why does even ProPhotoRGB(0,0,255) look blue at all on a monitor if it's so dark?

The answer is that ProPhotoRGB colors are converted to displayable colors. Taking the specific cases of displays that are close to sRGB and those close to Adobe RGB, here is what we get: Both displays will attempt to display RGB(0,0,255). An sRGB like display will show LAB=30,68,-112 while Adobe RGB wiil do LAB=30,69,-114. These are close because the "blue" of Adobe RGB and sRGB are the same. They are also well defined since ICC specifies the math to be used when converting matrix profile colorspaces.

But what color should a printer make? These are way beyond printable gamuts. Generally, profile software in Perceptual tries to preserve hue or hue and luminance. In the latter case the printer will print black. This is what causes the big differences between how profiles handle these colors. And a printer profile operates only on ICCLAB values.

My 9800 glossy profile converts ProPhoto(0,0,255) to  LAB(16,18,-56), a printable blue but much darker than what my monitor shows. Is that the "right" color? Who knows?

So this brings up the more general issue of what colors should we be working with. I generally put these in several categories from the narrowest set, those that the printer can print, to increasingly larger gamuts.

These are stimulus colors. Colors are perceived based on context. That is what stimulus colors surround them and what adaptation has occurred.

1. Colors that are actually printable. By definition, these colors are within the printer's gamut.
2. Colors that can exist, theoretically at least, on a surface. These are the set of all colors with completely arbitrary spectral reflectances, and are bounded by the MacAdam limits. They are more limited than all possible visible colors because they are formed by subtraction of wavelengths. In a sense, these might be considered the limits of a "perfect" printer with an infinite set of inks.
3. Colors that exist and can be created by mixing all possible wavelengths of light. These require emissive sources. Think mixing light from a large number of lasers, each with a slightly different visible wavelength and adjustable magnitudes.
4. Finally, there are the imaginary colors. These are colors with xy coordinates that are outside the stimulus gamut. They don't exist because there are no combinations of wavelengths at any magnitude that produce them. These are the ones outside the classic CIEXY horseshoe gamut most often encountered in ProPhoto RGB where the Green and Blues are imaginary.

When evaluating a profile, my criteria varies. It's most critical for in printer gamut colors. There shall be no visible banding or other anomalies for these. Next, are Mac Adam colors that are outside the printer's gamut but physically realizable. These should be rendered smoothly and be pleasing though they will not be accurate. Then there are colors that are real but not possible on a surface. How they are rendered is not something I care about too much though my preference is that hue be maintained. These, can't ever be printed but can appear in photographs with things like red LED Christmas lights . I expect the luminance and saturation to drop but the hue to remain close.

Andrew's profile stress test image is particularly useful as it contains a variety of all of these. The photos are near what printer gamuts can reproduce and the synthetics push things out further past MacAdam and imaginary color limits. He has a good video of this showing how to edit real, but saturated images pushing the printer's gamut to recover blocked colors.

By way of example I've attached a down sampled color coded mapping of Andrew's stress test image showing the colors that are in each of these 4 categories.

The color assignments are:
1. pale cyan: Colors inside the printer's gamut. This varies of course, I used my 9800 with an Epson Prem. Glossy for this.
2. blue: Colors that are physically realizable on a surface (within MacAdam limits) but not printable with my printer.
3. red: Colors that physically exist (within the human gamut) but only with emissive sources like lasers, leds, etc.
4. black: Imaginary colors. They don't exist but are useful mathematical constructs.


« Last Edit: December 27, 2017, 02:45:38 PM by Doug Gray »
Logged

Brad Paulson

  • Full Member
  • ***
  • Offline Offline
  • Posts: 205
Re: Choices Between Printer Color Profiling Engines
« Reply #39 on: December 27, 2017, 11:28:56 PM »

Attached are the complete charts.  Imperfectly lit but good enough for present purposes.  To observe the dark color gradations, you'll really need to click on the file and zoom in.  The darks are more colorful than they appear in the smaller thumbnails (and in person).  The i1P blues and gradations are distinctly better in person than the photos here might reveal, even in Perceptual.

Andrew, Doug, Graeme, Aaron and really all, thank you for your thoughts.  I'm quite grateful as I'm sure others are.  Doug, your last post I'm going to need to mull over a bit and play around in Lab before fully digesting and responding. 

For now, I want to get this benchmarking up for anyone interested in that to see.  I'm beginning to think the real answers here are more complicated than I had thought.  But for people working in ProPhoto space like me (so far), there's 1000 words in every picture.
Logged
Pages: 1 [2] 3   Go Up