What about printer managing color? Ctein says "yes."

[digitaldog]

If you know what idiosyncracies to avoid in the perceptual and relative rendering flavors of various ICC profiles (yes, even relcol should be considered another vendor biased flavor of rendering), and you have competent PS editing skills.

Here's the differences in RelCol using identical spectral data with two products (i1Profiler and Copra):

--------------------------------------------------

dE Report

Number of Samples: 988

Delta-E Formula dE2000

Overall - (988 colors)
--------------------------------------------------
Average dE:   0.92
    Max dE:   7.09
    Min dE:   0.02
 StdDev dE:   1.08

Best 90% - (888 colors)
--------------------------------------------------
Average dE:   0.61
    Max dE:   2.01
    Min dE:   0.02
 StdDev dE:   0.47

Worst 10% - (100 colors)
--------------------------------------------------
Average dE:   3.60
    Max dE:   7.09
    Min dE:   2.07
 StdDev dE:   1.28

--------------------------------------------------

935 patches converted without Dither in PS using the two profiles set for RelCol.

[MHMG]

Here's the differences in RelCol using identical spectral data with two products (i1Profiler and Copra):

Andrew, I not withstanding the one patch with dE = 7 variance, i agree those two profile apps have excellent agreement between them. Yet I see much greater variance in logo Colorful versus logo Classic in their relcol tags, not just there perceptual tags (which would be expected). I suspect many third party media vendors still use this older PM5 software due to it's more liberal licensing. I also think if you open this comparative anaysyis up to Canon, Datacolor, Monaco, and others profile making apps, you are going to find more of what I was referring to in the performance of the relcol tag. I could be wrong, but relcol definitely wasn't best choice for Ctein's Apollo Soyuz image. Logo Classic Perceptual was my much preferred choice if I had to fix Ctein's image using a PS manages color approach.

I just bought i1Pro2 and have not had a chance to put i1Profiler through it's paces yet. Maybe I will find something new, but for now, I'm very familiar with PM5's three rendering recipes, and I routinely alternate between Colorful and Classic depending on image, and in this regard, this give me four flavors to try, 2 perceptual an 2 relcol variations because the relcol tags don't produce the same rendering. Go figure.

best,
Mark

[r010159]

Andrew, I not withstanding the one patch with dE = 7 variance, i agree those two profile apps have excellent agreement between them. Yet I see much greater variance in logo Colorful versus logo Classic in their relcol tags, not just there perceptual tags (which would be expected). I suspect many third party media vendors still use this older PM5 software due to it's more liberal licensing. I also think if you open this comparative anaysyis up to Canon, Datacolor, Monaco, and others profile making apps, you are going to find more of what I was referring to in the performance of the relcol tag. I could be wrong, but relcol definitely wasn't best choice for Ctein's Apollo Soyuz image. Logo Classic Perceptual was my much preferred choice if I had to fix Ctein's image using a PS manages color approach.

I just bought i1Pro2 and have not had a chance to put i1Profiler through it's paces yet. Maybe I will find something new, but for now, I'm very familiar with PM5's three rendering recipes, and I routinely alternate between Colorful and Classic depending on image, and in this regard, this give me four flavors to try, 2 perceptual an d2 relcols because the relcol tags don't produce the same rendering. Go figure.

best,
Mark

"According to him, the fact that epson color controls are fixing his edited image's color errors in the night sky and that he can make easier edits to the file his way than with ICC profiles is indeed a fortunate outcome for him and for this challenging image. Hard to argue him, I guess. Still, I'm am very curious to know why from a technical perspective. "

Perhaps more realistically speaking, something gets converted to sRGB before the image is printed?

The following is just my limited knowledge. I think that with the error stats between the two profiles, adding the possible error stats of each profile itself, depending on which profile is used, together with a possible Delta-E of up to one for the i1Pro (Chromix statement), to me says that at least one of those profiles can be IMHO virtually unusable. But I may be wrong here. Still, this may be an issue on how the profiler was configured, as stated earlier.

As an aside, what is this infamous "blue turns purple"? What causes it? I have had this problem many times in the past.

Thank you!

Bob

PS Now I can see the advantage of a wide gamut display using the equivalent color space, aRGB, which also encompasses most of the actual printers gamut. But then some suggest ProPhoto.

[Jager]

I think one of the greatest misconceptions of color-managed printer workflows is that they will just... work.  Get yourself a nice monitor, profile it with a Spyder or an i1, use good printer profiles, pay attention to color management throughout the workflow, and you're good to go.

Didn't work for me.  For years I'd see variance between what I saw on the screen - on my dutifully profiled monitor - and what came out of the printer.  What you end up doing is, like Ctein, throwing darts at the wall.  Something eventually works.  Usually. 

The epiphany for me was nine months ago when I bought two things I had resisted for years and years:  An Eizo CG monitor; and a GTI viewing station.

I used Dell monitors for many years.  And then Apple.  Great panels, I thought.  Evaluating prints was done in, well, whatever decent light I could find.

What I discovered was that the "color-managed" workflow I had carefully applied for so many years... wasn't so much.  It was a revelation to find that you can, indeed, have a soft proof sitting there on your monitor and that what soon comes out of your printer is so close it's almost eerie.  Instead of throwing darts at a wall - "let's print this one and see if it's any better" - you make changes in your editing software.  You may or may not like what is staring back at you, at what the soft proof is telling you - Jeff Schewe in his book aptly called the soft proof option "make my print look like shit," or words to that effect.  But the job of a proper soft proof is to impose the same constraints on contrast and dMax and luminosity that that paper you've chosen is going to give you.  Its job is to be honest.

There are a bunch of crazy good monitors out there nowadays.  4K and retina and all that other good stuff.

Only, they ain't.  To my knowledge, there are only two monitors sold to consumers that will accurately render pretty much all of aRGB, at the luminance and contrast levels that paper will ultimately present:  the NEC Spectraview and the Eizo CG.

Ctein needs a new monitor.

[MHMG]

As an aside, what is this infamous "blue turns purple"? What causes it? I have had this problem many times in the past.

Take a look at the attached image. It is in sRGB colorspace, but it first started out as a LAB image when I created it. It is fully in gamut for the sRGB color space, so if you convert back to LAB in PS and vice versa, it's not going to change in image appearance. And it should reproduce reasonably well on just about any calibrated display since all modern ones can reasonably render sRGB colorspace these days.  I used the CIELAB color model to make the target. The goal was to create one single blue color of constant hue while varying lightness and chroma from black to very nearly white tonal gradient, all the while staying neatly within the sRGB color gamut. This target accomplishes that feat as far as the LAB color model is concerned. Mathematically in the LAB color space it has a constant hue. Yet to the human observer with reasonable color vision, the hue doesn't look the same as we go from the most vivid "cobalt blue" hue down to the lower chroma desaturated colors which inevitably have to occur when one heads toward pure white or pure black in any RGB color space. As the color desaturates it turns more purplish blue in appearance. Thus, in practice, vivid blue colors that are out of gamut or nearly out of gamut start getting remapped to lower chroma purplish blue colors as we attempt to remap this part of the color space into the printer color space.  The cause of this, IMHO, is that the CIE color model, while incredibly good at mimicking human color vision under controlled lighting conditions, is not perfect. The biggest errors occur in the blue sector of the color wheel. There are lesser errors in yellow-red as I recall, but the one that trips up faithful digital image reproduction more often than not is in the blue sector.

I personally believe the problem stems from the CIE committee's use of a simple cubic power function to model the lightness curve of the human visual system. It's a pretty good approximation that makes the math easier, but work done in the 1950s by Bartleson and Brenneman at Eastman Kodak showed a more nuanced set of human "brightness curves" and lateral adaptation to scene luminance values that deviate significantly from the simpler cubic equation. Most profiling packages attempt to "fix" the problem by various fudge factors used in the perceptual rendering tables, but IMHO, it could be more elegantly fixed by eventually updating the lightness power function in the CIE math. Yeah, I know, that will take a lot of time. Someone please give me a a nice institutional grant to go off into the wilderness for a couple of years and work on it.

Anyway, Ctein's Apollo Soyuz image is a great example of the blue turns purple issue in digital color reproduction (probably one of the very best examples I've seen) creeping into not only printer drivers and ICC profiles, but also very likely in the scanner output when the image was converted from film to digital. We expect the search light beams to be of a constant hue as they trail off into the night sky. That's what Ctein wanted to reproduce, not an unrealistic cyanish-blue to magenta blue transition that began in the file, and can either get amplified or nullified depending on the printer color conversion pathway. Ctein liked the nullified output that the Epson color controls gave him on his image. But print this target if you like using the Epson color controls.  I did. This target's cyanish blues and purplish blues should all be preserved if we want an accurate rendition of this target image. We should reproduce the color gradient exactly. However, the Epson color controls start to "fix" this hue gradient, presumably because Epson engineers who constructed the embedded LUTS in the Epson color controls decided that this hue shift more often than not represents an unwanted error in the original file (as it was in Ctein's image) rather than a truly intended color gradient (as it is in this target image).

cheers,
Mark
http://www.aardenburg-imaging.com

[r010159]

This is all excellent information! I have a question. how did you compress the color informatiin to be within the bounds of sRGB?

Bob

[Schewe]

Ctein needs a new monitor.

Yep...when you get a quality display, you can use soft proofing very well. Most of the people who argue that soft proofing doesn't work are using cheap-ass displays. It's really hard to use a display that barely has sRGB and expect soft proofing to work. I can do it in a pinch but I much prefer my NEC displays. Sadly many of the "experts" don't use quality displays and don't really believe soft proofing works. It does and does so very well.

[MHMG]

This is all excellent information! I have a question. how did you compress the color informatiin to be within the bounds of sRGB?

Bob

How I made that target is definitely not rocket science. A bit color geeky, but I believe I could teach the required steps to just about any one participating in this forum.  It's just that it has a fair number of steps, and requires an additional explanation on how to use a hue-slice tool I created for photoshop,  thus would require a more formal write-up then seems fitting for today's forum discussion. I've already been long-winded enough in my previous posts. I'm going to give you guys a break

best,
Mark
http://www.aardenburg-imaging.com

[digitaldog]

The following is just my limited knowledge. I think that with the error stats between the two profiles, adding the possible error stats of each profile itself, depending on which profile is used, together with a possible Delta-E of up to one for the i1Pro (Chromix statement), to me says that at least one of those profiles can be IMHO virtually unusable. But I may be wrong here. Still, this may be an issue on how the profiler was configured, as stated earlier.

Don't understand. There's no error per se, there are differences and in theory, a RelCol conversion should be the same. It's not.

As an aside, what is this infamous "blue turns purple"? What causes it? I have had this problem many times in the past.

A lot has to do with the warts in Lab color space. More modern profilers adjust for this in their color engines, better, more modern profilers should not show this issue (much or at all).

[robgo2]

Yep...when you get a quality display, you can use soft proofing very well. Most of the people who argue that soft proofing doesn't work are using cheap-ass displays. It's really hard to use a display that barely has sRGB and expect soft proofing to work. I can do it in a pinch but I much prefer my NEC displays. Sadly many of the "experts" don't use quality displays and don't really believe soft proofing works. It does and does so very well.

Speaking of displays, I am using a 23" Apple Cinema Display, which is about 8 years old.  What are the best current displays for the purpose of editing photos?  Are the new high resolution displays unsuitable for reasons of limited color space?  Is a new generation right around the corner?

Rob

[Schewe]

What are the best current displays for the purpose of editing photos?

I use NEC and I successfully soft proof with it.

[MHMG]

I use NEC and I successfully soft proof with it.

+1 NEC Spectraview

[digitaldog]

I use NEC and I successfully soft proof with it.

+2, for mean years. It's a critical component for successful soft proofing. That and calibrating to that goal which I'm not sure is happening in this case.

[howardm]

+3 NEC Spectraview

[donbga]

+3 NEC Spectraview

+4! NEC wide gamut monitor with NEC Spectraview colorimeter.

This combination never lets me down unless I do something stupid in soft proofing.

I should also point out that I don't print client but make prints for myself. I seldom expect perfect color mapping. When skin tones look acceptable other slight color variations in clothing, etc. don't upset me.

Skies with heavy gradients can be upsetting so I go back to PS and back off on the density of gradient layer and that usually makes me happy. 99.9% of the time people viewing a print won't notice slight color errors anyway.


Don Bryant

[deanwork]

+4 Spectraview

I've used the 27" Nec display with Spectraview  software for about 5 years now and have found it excellent and very consistent and a great value.

I make all my icc profiles with the i 1 Pro 2 X-Rite package.

What I have found over the years is that the soft proofing for all the gloss and sen-gloss papers I use, whether fiber based or rc is really excellent from my display to the hard proof, and that includes the hue and tone of the paper base.

 What I have found not so perfect is the paper base luminance on the screen when using matte media, like Canson Rag Photographique, Hahnemuhle rag media, as well as the coated kozo type papers or various fabrics. For me the soft proof on the screen is falling somewhere between having the paper white box checked on and having it checked off. I find the soft proof looks a little flat in the paper base with the paper white box checked on and too bright with it unchecked. So, I'm thinking somewhere between the two when using most matte media on all my printers. But it is definitely better in all cases using soft proofing than not using it at all! I am reluctant to share my profiles with clients who have poorly calibrated or cheaper displays. Even on my Apple displays I find soft  proofing on matte media quite a bit less accurate than with the NEC.

john

+3 NEC Spectraview

[deanwork]

+5 Spectraview

I've used the 27" Nec display with Spectraview  software for about 5 years now and have found it excellent and very consistent and a great value.

I make all my icc profiles with the i 1 Pro 2 X-Rite package.

What I have found over the years is that the soft proofing for all the gloss and semi-gloss papers I use, whether fiber based or rc is really excellent from my display to the hard proof, and that includes the hue and tone of the paper base.

 What I have found not so perfect is the paper base luminance on the screen when using matte media, like Canson Rag Photographique, Hahnemuhle rag media, as well as the coated kozo type papers or various fabrics. For me the soft proof on the screen is falling somewhere between having the paper white box checked on and having it checked off. I find the soft proof looks a little flat in the paper base with the paper white box checked on and too bright with it unchecked. So, I'm thinking somewhere between the two when using most matte media on all my printers. But it is definitely better in all cases using soft proofing than not using it at all! I am reluctant to share my profiles with clients who have poorly calibrated or cheaper displays. Even on my Apple displays I find soft  proofing on matte media quite a bit less accurate than with the NEC.

I use somewhere in the realm of 2300 patches for the profiles.

john

[PeterAit]

Yep...when you get a quality display, you can use soft proofing very well. Most of the people who argue that soft proofing doesn't work are using cheap-ass displays. It's really hard to use a display that barely has sRGB and expect soft proofing to work. I can do it in a pinch but I much prefer my NEC displays. Sadly many of the "experts" don't use quality displays and don't really believe soft proofing works. It does and does so very well.

I agree with Jeff. Four years ago I got 2 NEC wide-gamut displays, Spectraview software, and a good puck. Within half a day I had my Epson 4880 (and now 7900) prints matching my display very closely and it has stayed that way ever since. I almost never have to make more that one print of any image to get it right, and I cannot imagine printing without softproofing.

I will also point out that this Ctien first had one opinion, then reversed it, and now seems to have re-reversed it. This tells me that whatever print differences he is talking about are subtle and unimportant, otherwise why would he be waffling?

[Jager]

Speaking of displays, I am using a 23" Apple Cinema Display, which is about 8 years old.  What are the best current displays for the purpose of editing photos?  Are the new high resolution displays unsuitable for reasons of limited color space?  Is a new generation right around the corner?

Rob

There you go, Rob.  Lots of NEC Spectraview users here, so if you get that one you'll have a nice support group.  The only other panel that a serious printmaker would want to consider is an Eizo CG screen.  Either will get you where you need to be.  To my knowledge, that's it.  There aren't any others.

Which isn't to say there aren't a bunch of really nice monitors out there.  The problem is they're all designed for mainstream consumers.  Video gaming.  Watching movies/video.  Doing web stuff.  They're sRGB.  They're built for high-contrast and high-brightness, the very things that sell a consumer when he walks into Best Buy or the Apple Store.  And the latest extra premium models are ultra HD.

All of which is what you want when you're watching that Blu-Ray disk of Gladiator.  Or the latest CGI flick.

And they'll work fine for screen-based photo editing.  Which, let's be honest, is where most images end up these days.

But if your ultimate destination is paper, they will, at least part of the time, leave you wondering "what the hell happened?"  Like Ctein.

Last thing is what Andrew mentioned ("That and calibrating to that goal which I'm not sure is happening in this case.").  Color-profiling a NEC or an Eizo isn't enough.  You have to calibrate your monitor for paper.  Luminance and contrast have to come way down.  So much that it hurts.  I don't use my Eizo for anything other than printing.  It's too dismal for anything else.

The flip side is, well, printing is glorious. 

[robgo2]

Thanks for all the replies regarding displays.  Almost everyone here seems to use NECs rather than Eizos.  Is there a reason for this other than price?

Rob