Dan Margulis Applied Color Theory

Why Does Blue Turn Purple?
   
Date: Tue, 1 Apr 2003 19:31:34 +1000
   From: Stephen Marsh
Subject: Blue turns purple...

I have a question that has been bugging me for a while now, on the infamous blue turns purple issue. Why is the end user left to fix this, it is a known issue and the profile seems a better place to address this than with post separation edits (or pre sep edits). I hope that those with profile measuring/generation/editing experience may know the answer.

Many common separation methods have issues with out of gamut colours such as rich blue - often the final conversion will have hue values near 100c90m, when a human would probably choose to have the basic hue reading around 100c70-75m.

So why don't people who generate profiles limit the max build to have lesser magenta? Or to generate two profiles and have this as an option if it does not sound appealing as a default.

If the colour is out of gamut one can't argue about the resulting colour, but blue would be better than purple.

http: //www.brucelindbloom.com/MunsellCalcHelp.html#BluePurple

Sincerely,

Stephen Marsh.
________________________________________________________________________

Date: Tue, 01 Apr 2003 06:53:07 -0700
   From: Andrew Rodney
Subject: Re: Blue turns purple...

Blue1s going purple do happen and it was once more a problem than I see it today, at least with the package I use to build profiles (ProfileMaker Pro 4.x). My most saturated blues in a spectral gradient (going to CMYK or RGB output devices) remain nice and pure blue. Its one of the reasons I like this package so much. I also find that you can get hosed when doing blind conversions and specifying a Perceptual Rendering intent. There1s no spec for how this mapping is done so every profile package has their idea of how to do this. But I do see this issue cropping up from time to time and I would examine the methods, instruments and profile package used to generate the profile.

Andrew Rodney
________________________________________________________________________

Date: Tue, 01 Apr 2003 12:31:23 -0600
   From: "R. Lutz"
Subject: Re: Blue turns purple...

Here is another link to add to your collection on this topic.  It is from an interview of Robin Meyers by Jeff Harmon.  Perhaps a choice other than L*a*b* for the PCS is what is needed. http://www.colorhythm.com/s_resources/robinmeyers_2.html

________________________________________________________________________

Date: Tue, 1 Apr 2003 15:09:45 EST
   From: Dan Margulis
Subject: Re: Blue turns purple...

Stephen writes,

I have a question that has been bugging me for a while now, on the infamous
blue turns purple issue. Why is the end user left to fix this, it is a known
issue and the profile seems a better place to address this than with post
separation edits (or pre sep edits).

Because in most cases the cure would be worse than the disease.

Conventional CMYK isn't just slightly worse than RGB at producing blues--it's drastically, horribly worse. There isn't any separation method that can cater to every possibility.  The following are the basic options.

1) Allow the out-of-gamut blues to go purple and allow in-gamut blues to take their natural value.

2) Force the OOG blues to be blue by subtracting magenta but do nothing to less saturated blues, meaning there will be no difference between blues that were originally OOG and blues that are substantially less pure.

3) As above, except tone down the less saturated blues somewhat, along the lines of a "Perceptual" rendering intent but stronger, to retain some slight distinction between purer and duller blues.

4) As above, except smash less saturated blues until they are almost gray.

5) As above, except drive any part of the image that isn't brilliant blue towards yellow.

6-9) As 2-5, except this time not only subtract magenta from the blues, but also add black, so as to attempt to match the luminosity of the original RGB blue but not its purity.

Obviously, one can come up with images that favor any of these nine approaches. But let's try to do a realistic analysis of which one we would choose for everyday work. First, let's suppose that our only possible choice is between #1 and #2. Here are the possible images we might face.

A) No OOG blues at all, or totally irrelevant to the image. In that case, no advantage to either #1 or #2.

B) OOG blues exist, but it's only important that they be blue, not detailed, not necessarily differentiated from other blues. Those who have attended my class will remember "Interior at Night" as such an image. In that case, clear advantage to Method #2.

C) OOG blues exist AND either have to be detailed OR differentiated from other blues. In that case post-conversion correction will be needed because neither #1 nor #2 is satisfactory as is. And in that case #1 is very much better--with #2 we'll be in Selectionsville because we have to differentiate blues that no longer are differentiated.

In my experience, we're about 50 times as likely to find images of Type C than Type B. So there's basically no case at all for using Method #2. The real choice is between Method #1 and something more devious, like Method #3. So, let's go through it again. This time, the types are:

D) No OOG blues at all, or totally irrelevant to the image, *and no other significant blues either*. In that case, no advantage to either #1 or #3.

E) As above, except there are other blues in the image. In that case, strong advantage to #1, because #3 will tone them down unnecessarily.

F) Important OOG blues that will be managed reasonably well by #3 but not by #1.

G) Important OOG blues that will not be managed well by either. In that case it's not evident that there's an advantage to either: #3 may get you closer but #1 is likely to be easier to correct.

This time, the question is whether we more commonly find Type E or Type F, and obviously it's Type E, although by no means 50-1. However, it's more complicated than that.  If you have Type E, Method #3 won't do as well as Method #1 but it won't be a tragedy either. But if you have Type F, Method #1 is going to be *much* worse than Method #3.

So it may boil down to what kind of images you're dealing with. For me, Type E is way more common than Type F. A sports photographer taking pictures of teams with blue uniforms would be on the opposite side of the fence and would probably need to use Method #3.

So why don't people who generate profiles limit the max build to have
lesser magenta? Or to generate two profiles and have this as an option if it
does not sound appealing as a default.

You could definitely have two or more profiles but the question would be how much effort it would be worth, which would depend upon how often the problem comes up. Presumably one of these false profiles would be method #4 above, for which you'd need either a dedicated profile editor or a Photoshop development team more disposed to put profile editing in their product.

In many ways this is the same subject recently brought up by Richard Chang, whose problem was an overly optimistic profile that blew out highlights and plugged shadows in the interest of more midtone contrast.  Such a method is quite traditional, as is the method of allowing blues to go purple. The reason that these methods are traditional is that they work well--most of the time.

Unfortunately, the law of separation is: the better a method does on typical images, the worse it will do when something exceptional comes along. So, you have your images with blues that are purple and Richard has his with blown-out highlights.


Dan Margulis
________________________________________________________________________

Date: Tue, 1 Apr 2003 11:43:23 -0800
   From: Jan Steinman
Subject: Re: Blue turns purple...

  From: Stephen Marsh

I have a question that has been bugging me for a while now, on the infamous
blue turns purple issue.

Trade you problems! I have a client who wants "periwinkle," but it seems I'm giving her something that is too blue (from a color laser).

Have you tried using relative colorimetric rather than perceptual conversion? (or vice-versa?)

--
: Jan Steinman -- nature Transography(TM): <http: //www.Bytesmiths.com>
: Bytesmiths -- artists' services: <http: //www.Bytesmiths.com/Services>
________________________________________________________________________

Date: Tue, 1 Apr 2003 17:28:51 -0500
   From: Annette Murray
Subject: RE: Blue turns purple...

And your laydown order on press will have an impact on blues also. Cyan over Magenta vs. Magenta over Cyan.

Annette Murray
Prepress and  Color Consultant
______________________
ANRO Inc.
222 Lancaster Avenue
Devon, PA 19333
800.355.2676
610.687.1200 x241
FAX: 610-254-8539
http://www.ANRO.com
________________________________________________________________________

And only a few months later, a photographer reopens the topic.
________________________________________________________________________

Date: Sat, 27 Sep 2003 08:00:17 -0700
   From: Doug Walker
Subject: Magenta blues

Dear Group,

Can anyone shed some light on magenta blues?

In my color managed workflow using Gretag EyeOne Display and Apple  Studio Display I just recently spent considerable time properly destining a highly blue and pink image destined for output as a catalog cover at a bucket shop.

The press uses the Web Coated SWOP v2 as a guide and are not color managed.

Soft-proofing revealed much out of gamut and so the colors were reigned in, adjusted in 16bit and then converted to Web Coated SWOP v2

The actual proofing provided my clieint looked very close.

However, the printed cover is very magenta even though all sampled colors were in Gamut and printable.

Is there a magic cutoff combination that usually yields a magenta-ish blue that one can quickly sample via the numbers?

Thanks in advance,

Doug Walker, FP
"Specializing in People on Location in a Clean, Bold Classic Style!"
website: http://www.walkerphoto.com
Member, ASMP, APA SF
________________________________________________________________________

Date: Sat, 27 Sep 2003 17:52:47 -0500
   From: David Riecks
Subject: Re: Magenta blues

Doug:

I find it very useful to use the softproofing feature in photoshop to see how the colors will render in SWOP. In addition I check the actual CMYK values using the Agfa "Postscript Process Color Guide" to see what that combination of inks will produce. Over time I've trained my eyes to see things that were always there, but I would consistently miss. "Magenta blues" are one of those items.

It's not perfect, but looking up the values in the book, you will quickly discover what you thought on screen to be a deep blue is often a blue-purple.

If you have a deep dark blue, first off, you've got to figure it's going to have very little yellow (I know, duh), but for something with 90 percent cyan, and 5 percent yellow, I find that it looks pretty blue till you start getting in the 60 to 70 percent range on the magenta. 90c50m5y looks nice and blue, but 90c80m5y looks "very purple" in my book (and I mean that literally). ;-)

If you want to purchase this book you won't find it on Amazon, so go to http://support.agfa.com/support/EN/shopListItemsHome.jsp choose your country (seems stupid cause I only see US on mine), then choose "books" from the line of items in red that appear. Scroll down to the "T's" because the consider the title to begin with T as it's titled "THE Postscript Process Color Guide" ;-)

Don't even think of using their document "find" box as that will tell you that ZERO documents exist for the search term!!! (don't even get me started).

If you do a search on Amazon, you will find a book by Michael & Pat Rogondino which does contain a lot of colors as well, but I don't like the way it's laid out. It's much harder to find the right color swatch IMHO.

Hope that helps.

David

David Riecks (that's "i" before "e", but the "e" is silent)
 http://www.riecks.com/
Midwest/Chicago ASMP  *   ph/fax 877-646-5375
________________________________________________________________________

Date: Sun, 28 Sep 2003 00:11:28 -0000
   From: Stephen Marsh
Subject: Re: Magenta blues
 
Is there a magic cutoff combination that usually yields a magenta-ish
blue that one can quickly sample via the numbers?

Doug, this is the problem with most profiles and sep methods that I have come over, when it comes to out of gamut high saturation blues. If more press/proofing output profiles used XYZ as the PCS, then we would not have this problem every time we separate a saturated blue (or red in some cases turning orange). It seems that XYZ based profiles are rare, when compared to all the LAB based ones out there in common circulation.

I asked a question on the topic, after getting irritated at having to correct yet another blue sep, Dan gave a very good answer in colour theory terms on the trade-offs of limiting a profile to get around the issue. Perhaps one needs two profiles if using LAB based profiles, one for most work and one for when saturated CMYK blue is critical.

It all depends on what 'blue' we are talking about - and the ink and stock and other variables. Tolerances can be small for a large amount of variation and it can really help to know how the target conditions are if this is a critical colour situation.

For some general 'high saturation' ratio's in SWOP type conditions (flatsheet should be similar), perhaps consider mixes like -

100c 75-70m ?y ?k - with minor amounts (5-10) of the two opposing inks, but one must be careful with high magenta values in blues (60>).

100c 60-55m to play things a bit safer.

100c 50m if you really are not sure.

Swatches are one thing, a natural image another - various tonal and colour transitions will play a bit part in the perception of the colour.

P.S. If you can design a job with yellow elements around the blue (with little Y in the blue) - then you can attempt to make the weak CMYK blue appear a little more saturated (trapping may be needed).

Regards,

Stephen Marsh.
________________________________________________________________________

Date: Sat, 27 Sep 2003 23:21:01 EDT
   From: Dan Margulis
Subject: Re: Magenta blues

Doug writes,

In my color managed workflow using Gretag EyeOne Display and Apple
Studio Display I just recently spent considerable time properly
destining a highly blue and pink image destined for output as a catalog
cover at a bucket shop....Soft-proofing revealed much out of gamut and so the
colors were reigned in, adjusted in 16bit and then converted to Web Coated
SWOP v2. The actual proofing provided my clieint looked very close. However, the
printed cover is very magenta even though all sampled  colors were in Gamut
and printable.

The SWOP Web Coated v.2 profile is the likely culprit--its rendering of deep blues is considerably more purple than found in other profiles such as Photoshop's traditional engine.

When taking deep blues from RGB, particularly Adobe RGB or something wider-gamut, into CMYK there are just too many holes in the dike--no profile can cater to every situation. With this particular profile, you're likely to get better differentiation in the blues than with others. The price you have to pay for it is this too-purple business from time to time.

Deep blues on press are a real land mine for color management.  It's very common on press for the magenta ink to run hotter one day than it did the last, whereas an overly green cyan ink is rare. Plus, a deep blue that's too cyan is likely to be much more acceptable than one that's too purple.

So, for critical blues, it wouldn't matter to me if I had a profile and affidavits from Jim Rich and Chris Murphy that they'd measured the blues as being accurate on the same press and the same paper I was printing on. I'd still undershoot the magenta by three or four points.

Dan Margulis
________________________________________________________________________

Date: Sun, 28 Sep 2003 14:52:32 -0400
   From: Jim Rich
Subject: Re: Re: Magenta blues

Dan,
 
As for me proclaiming that I give out affidavits that a measuring technique is flawed or that color management is perfect... then you might have me confused with someone else. In the recent past, I have just stated that I have found color management works in a lot (not all) situtations.

And your observation about deep blues just goes to show that color managment has a few holes in it. The point you  bring up is more than valid. You still need color skill to make color work at all stages of the imaging process.

Best Regards,

Jim Rich
________________________________________________________________________

Date: Mon, 29 Sep 2003 09:39:02 EDT
   From: Dan Margulis
Subject: Re: Magenta blues

Jim Rich writes,

As for me proclaiming that I give out affidavits that a measuring technique
is flawed or that color management is perfect... then you might have me
confused with someone else. In the recent past, I have just stated that I have
found color management works in a lot (not all) situtations.

I think you have *me* confused with somebody else. I suggested facetiously that you might issue an affidavit that a certain press printed a certain combination of inks on a certain paper as blue at a certain time. This is a long way from an affidavit that color management is perfect or anything else like it.

If I did the measurement myself, I would certainly be happy to issue such an affidavit, but I would caution people that because deep blues are such a problem area in CMYK, one shouldn't necessarily create a profile that tried to accommodate my measurement, and one shouldn't necessarily create files that cater to the measurement.

And your observation about deep blues just goes to show that color
managment has a few holes in it. The point you  bring up is more than valid. You still
need color skill to make color work at all stages of the imaging process.

Correct.

Dan Margulis
________________________________________________________________________

Date: Mon, 29 Sep 2003 10:22:16 -0400
   From: Lee Clawson
Subject: Re: Magenta blues

Doug,

a highly blue and pink image destined for output as a catalog
cover at a bucket shop. However, the printed cover is very magenta even
though all sampled colors were in Gamut and printable.

What did the printer say??

I've had this occur even when the colors are pastel range. Blue opposing Magenta/pink seems hard to balance on a press. Usually I hear that the M ran slightly hot. I'm told paper (cover stock) has big effect. Most times I'll reduce M in blue by 5% and as much as 10%.

Lee

P.S. I agree with Dan...
...a deep blue that's too cyan is likely to be much more acceptable than one
that's too purple.
________________________________________________________________________

 Date: Mon, 29 Sep 2003 08:00:04 -0600
   From: Andrew Rodney
Subject: Re: Magenta blues

on 9/28/03 12:52 PM, Jim Rich wrote:

And your observation about deep blues just goes to show that color managment
has a few holes in it. The point you  bring up is more than valid. You still
need color skill to make color work at all stages of the imaging process.

It1s somewhat nonsense. Yes I1ve seen plenty of ICC profiles (and Sep tables) produce magenta or cyan shifts to blues. My profiles don1t do this (my package makes very clean blues from the most saturated spectral gradient). Good profiles make very good blue conversions. Bad profiles (and other conversion methods) don1t. I1d be willing to take a good RGB test image anyone has, build a custom profile for a proofing device and do the conversion next to any method Dan wants to try using the same RGB file to illustrate this.

Andrew Rodney
http://www.imagingrevue.co
________________________________________________________________________

Date: Mon, 29 Sep 2003 17:42:33 -0000
   From: "David Bishop"
Subject: Re: Magenta blues

However, the printed cover is very magenta even though all sampled
colors were in Gamut and printable.

One thing not mentioned in other comments to your question is the actual ink/press/
paper combination on press and their variations.

First of all, no matter what you do, whether it's soft proofing to your desktop ink-jet or Matchprinting off a Scitex proofing device you're still making dyes or pigments other than actual ink try to simulate what will happen on press. It can be very close, certainly close enough for clients to sign-off on and still be satisfied when the sheet comes off the press. BUT

How do you control the printer, the relative moisture of the paper, the viscosity of the inks or what's booked along with your image before yours? I think your problem lies in that direction.

 For all your efforts there still needs to be someone on press to tell them 'back off the magenta over here, a bit more cyan here.....' to get that last bit of correction that will satisfy the client.
 
One option, if you're willing to take the time, is to ask the printer for a digital CMYK file that has a similar color (magenta/blue in your case) so that you can compare CMYK values and see what the differences are. Ideally a match print of the file under a 5500?K light source will help you tweak your values to match what the press is capable of.

david/sf

Adobe Photoshop training classes are taught in the US by Sterling Ledet & Associates, Inc.