Perceptual Gamut Compression? Little if any in PM5 and I1Profiler

[Doug Gray]

Here's Adobe's submission. It's now in ISO 18619
http://www.color.org/BlackPointCompensation.pdf

If implemented, it should be optionally done by the CME. It's not a part of the ICC spec.

It's main use these days is when using Relative Intent when the source image has materially darker grays than the destination profile. Proper RI tables in both V2 and V4 profiles will clip at the media's black density limit. Some vendors had done BPC in the RI tables in the past but this is prohibited in V4 and strongly discouraged except for in house use for V2.

Perceptual Intent seems to have always implemented BPC so selecting that in PI does little if anything. At least I've never run across a PI that didn't.

There is a minor weirdness. V4 profiles force a BP at approximately L=3 to match the standard output reference medium but in the PI only. There is a discussion of this in the FAQ. I1Profiler's V4 implements this small clip in the Curves front end before the LUTs. Photoshop ignores it so you get a full BPC like in V2 PIs. LCMS doesn't. It's a nearly invisible effect.

[GWGill]

Perceptual Intent seems to have always implemented BPC so selecting that in PI does little if anything. At least I've never run across a PI that didn't.

Unlike the V4 spec., nothing in the V2 spec. says that P/SI has to map to/from a reference black point. Some particular vendors may have been doing this forever in their V2 profiles, but they seem to have kept that bit of "secret sauce" to themselves. So BPC does have uses for R.Colorimetric, Perceptual and Saturation profiles, depending on where they come from.

[Doug Gray]

Unlike the V4 spec., nothing in the V2 spec. says that P/SI has to map to/from a reference black point. Some particular vendors may have been doing this forever in their V2 profiles, but they seem to have kept that bit of "secret sauce" to themselves. So BPC does have uses for R.Colorimetric, Perceptual and Saturation profiles, depending on where they come from.

There really are no conditions that apply to PI. At least it V2. There could be and probably are PI profiles that might need BPC. I just haven't run across them outside of the L=3 clip in V4 which is ignored by Photoshop. Might even be of some use to me in Photoshop if they get around to being V4 compliant.

[Doug Gray]

While this topic is mostly about gamut clipping for the highly saturated colors printers can print but that are outside sRGB and aRGB there is a small difference between RI and PI in the way colors that are not printable are mapped into printable ones.

Here's a chart, in sRGB, of color gradients which are all in sRGB but the left hand saturated side is not printable. It shows what colors are printed instead for Perceptual Intent as well as Relative Intent. The requested color starts on the left with the unprintable color and gradually desaturates, keeping the same hue, to a fully desaturated gray on the right. In each group of 3 gradients, the first gradient is the requested color. The second is the color rendered with Relative Colorimetric and the third is rendered with Perceptual Intent.

The printer was a Canon 9500 II with a Baryta semigloss, no OBs and an I1Profiler profile. Default settings, V4, large LUTs.

You can download this image and soft proof to your printer and profile to see how closely these results match a different setup.

Some observations:

Notice the well known blue to magenta hue shift as the chrominance drops on the middle set on the right hand side. This effect is due to an apparent hue shift in certain colors even though the actual Lab hue is constant.

The major difference I see between the PI and RI stripes compared to the image stripe is that the PI portions that are out of gamut seem slightly closer visually than the RI even though the RI dE is somewhat greater. That's clearly a conscious decision to map colors for a better perceptual map at the expense of a higher dE.

[Steve Upton]

Notice the well known blue to magenta hue shift as the chrominance drops on the middle set on the right hand side. This effect is due to an apparent hue shift in certain colors even though the actual Lab hue is constant.

Can you explain what you mean in a little more detail?

"actual Lab hue" being constant in blues would mean that it wouldn't appear to be constant due to Lab warping (perceptually) in the blues.

[Doug Gray]

Can you explain what you mean in a little more detail?

"actual Lab hue" being constant in blues would mean that it wouldn't appear to be constant due to Lab warping (perceptually) in the blues.

Exactly. That was what I was noting but it was just an observation unrelated to printing or Perceptual or Relative Intent. The top stripe of each 3 stripe set is constant Lab hue which means a/b is constant. To best see the blue->purple shift block out the lower two stripes in the group of three.

[pfigen]

I have both applications being discussed here, and to tell you the truth, neither is worth a crap for Perceptual rendering. I have used Gretag's Perceptual rendering for certain files going to offset press - one that I particularly remember was a lime green graphic on the side of a jet ski. Went completely flat under Relative but was fully delineated under Perceptual, but because Perceptual fucked up the entire rest of the image, I did what I almost always do when needing Perceptual for offset - make two conversions and layer the Perceptual on top of the Relative and paint it in just where it's needed. Worked like a charm and printed exactly as expected with the printing house wondering how I did what I did. This is not such a big deal for inkjet as the gamuts are generally much larger than offset, but it's still too bad that you basically can't do that anymore.

As far as Relative Colorimetric, actual printed output always looked just a RCH better from ProfileMaker as compared to Profiler, but most people would never ever see the difference in the real world. And, of course, Gretag had/has hands down the best profile editor I've ever used.

[Doug Gray]

I have both applications being discussed here, and to tell you the truth, neither is worth a crap for Perceptual rendering. I have used Gretag's Perceptual rendering for certain files going to offset press - one that I particularly remember was a lime green graphic on the side of a jet ski. Went completely flat under Relative but was fully delineated under Perceptual, but because Perceptual fucked up the entire rest of the image, I did what I almost always do when needing Perceptual for offset - make two conversions and layer the Perceptual on top of the Relative and paint it in just where it's needed. Worked like a charm and printed exactly as expected with the printing house wondering how I did what I did. This is not such a big deal for inkjet as the gamuts are generally much larger than offset, but it's still too bad that you basically can't do that anymore.

As far as Relative Colorimetric, actual printed output always looked just a RCH better from ProfileMaker as compared to Profiler, but most people would never ever see the difference in the real world. And, of course, Gretag had/has hands down the best profile editor I've ever used.

Profile editing is not for the faint of heart. The biggest risk is messing up the relationship of the AtoB and BtoA tables which should be the inverse of each other. This then affects softproofing. I played a bit with it but made more of a mess than anything useful. Glad you were able to get it to work to your satisfaction.

[digitaldog]

Profile editing is not for the faint of heart. The biggest risk is messing up the relationship of the AtoB and BtoA tables which should be the inverse of each other. This then affects softproofing. I played a bit with it but made more of a mess than anything useful.

It was pretty darn easy using Photoshop and Kodak's long gone Custom Color ICC. I still run it from time to time on a very, very old Mac. And yes, you can edit either or both table so there's no mismatch in what you get and what you see. Sometimes profiles only need one table edited, sometimes both.


As for GMB's profile editor, it was a huge mess of a GUI and one of the most incomprehensible products I've ever used. Editing a profile by editing images in Photoshop? Super easy.


Anyone here old enough to remember editing profiles using ColorBlind Edit? Pretty good too, not close to what Kodak provided however. What toolset for editing colors to affect a profile is more powerful and well know as Photoshop? I begged Thomas Knoll often to allow us to use PS to do this, never happened.

[Doug Gray]

It was pretty darn easy using Photoshop and Kodak's long gone Custom Color ICC. I still run it from time to time on a very, very old Mac. And yes, you can edit either or both table so there's no mismatch in what you get and what you see. Sometimes profiles only need one table edited, sometimes both.


As for GMB's profile editor, it was a huge mess of a GUI and one of the most incomprehensible products I've ever used. Editing a profile by editing images in Photoshop? Super easy.


Anyone here old enough to remember editing profiles using ColorBlind Edit? Pretty good too, not close to what Kodak provided however. What toolset for editing colors to affect a profile is more powerful and well know as Photoshop? I begged Thomas Knoll often to allow us to use PS to do this, never happened.

Sounds interesting. I'd love to see that as well.

I've edited profiles in Matlab. When you load them you have access to all the internal tags including luts and curves. I've used it to modify matrix profiles but only to change gamma to one on a ProPhoto profile. I think there is a way to do this in Photoshop too. I keep intending to make a B&W printer profile. Someday.

[hugowolf]

Eric Chan, who is now listed as second only to Thomas Knoll, on the CS6 splash screen, used to make ABW profiles for Epson 3800 and 3880 printers, until Apple decided to 'grayout' that option. Too much secret sauce.

I am old enough to think of Adobe as a font company, with a side line in a photo editing app. And then there were acquisitions.

Brian A

[pfigen]

Oh Andrew. Too bad you never got the hang of the Gretag tool. It really is a much better and way more professional product and let you edit the Absolute Colorimetric white points for ultra fine tuning proofing profiles for offset. The interface is perfect and super easy to understand. The old Kodak plugin, well, that's the one that's the mess and that you can only use Ps commands really limits it. I have that too but I never even think about using it. For anyone interested in serious profile editing, Gretag is really one of the only ways to go. I used to use it far more often than I do now, but I'm sure glad it's there when I need it.

For the record, Andrew has been giving the Gretag module a bad rap for years and for some reason he has not been able to grasp it. It's one of those products that once it clicks in, you wonder why you'd ever use anything else, especially for some of that hard to do stuff that nothing else can do. (and I think that Argyl has an editor but I don't know much about it, so maybe that's an alternative)

[digitaldog]

For the record, Andrew has been giving the Gretag module a bad rap for years and for some reason he has not been able to grasp it.

One of the worst examples of a workflow and GUI ever created IMHO. If you love it, great. I'll take ColorBlind Edit which predates it by years, or better Custom Color ICC any day of the week (and both can produce the edits of Absolute Colorimetric white points).

[digitaldog]

Eric Chan, who is now listed as second only to Thomas Knoll, on the CS6 splash screen, used to make ABW profiles for Epson 3800 and 3880 printers, until Apple decided to 'grayout' that option.

Can you explain that bit about a grayed out option?

[hugowolf]

Can you explain that bit about a grayed out option?

Under all reccent versions of OS X, there is no way of applying a secondary profile when in ABW mode.

Brian A

[digitaldog]

Under all reccent versions of OS X, there is no way of applying a secondary profile when in ABW mode.

Where was that? In the Epson driver/Color Handling?

[hugowolf]

Where was that? In the Epson driver/Color Handling?

With Eric's profiles, you would set the printer for ABW mode, effectively setting the printer to manage color, then in the printing application you would set it to Photoshop (or Lightroom, etc) manages color and select Eric's profile for that specific paper. This can still be done in Windows, but on a Mac you lose the ability to set both driver and the printing app manage color. If you set Lithroom to mamage color, you cannot get to the driver's ABW mode.

http://people.csail.mit.edu/ericchan/dp/Epson3880/abwprofiles.html

Brian A

[Doug Gray]

You can work around the Apple limitation by:

1. Converting the B&W image to Eric's profile using the desired intent. This runs the RGB values through Eric's curve settings.
2. Assign the image to the sRGB profile.

Let the printer "handle color" then print it using ABW.

Eric's B&W profiles incorporate BPC across all intents but he maps AtoBn curves to their actual values which means the view proof works correctly.

[hugowolf]

I am running Windows, so it is no prolem for me. Canon ipf accounting software is Win only, Qimage is Win only. Quickbooks used to be Win only.

Erics profiles are also only going to last the life of a 3880, then they are gone. Good while they lasted.

Brian A