Adobe Photoshop Training with Dan Margulis

Greetings to the group:

I am growing quite interested in the possibilities with Lab curve moves. Thanks to Dan's articles and book, I find myself converting files to Lab and back for various color shift and saturation solutions. However, with all the charts and articles, I still find it difficult to take the magic out of the "a" and "b" channels. I also find it confusing that when these channels are manipulated, they're often described as red, blue and green. Then when you go from RGB to CMYK to Lab and back, I'm lost as to what really happened.

CMYK, for example, makes alot of sense to me. I basically know what channel to go to and what quarter-tone, half-tone, etc, to use, what it does and how it relates to RGB.

Experience, I'm growing to understand, is the best teacher. However, sometimes someone mentions a simple word, analogy or sentence that hits the nail on the head for me. Does anyone have a straight-forward way of taking the mystery out of "a" and "b"?

--

Eric Curtis M. Basir (Bond)

Photo Grafix

http://www.abetterreality.net

(847) 673-7043

___________________________________________________________________________

My initial comment is that if your output is going to be RGB, it's generally not good to convert to CMYK and then back to RGB, even if you use LAB as an intermediary.

As to LAB, the A channel will (generally speaking) shift the color from Red to Green and vice versa, and the B channel will shift the color between Blue and Yellow. This comes in handy for dealing with color casts. Re-read Dan's articles and book for more sophisticated uses - I'm still re-reading myself.

Maris

___________________________________________________________________________

Hi Eric,

I would suggest a little experiment that may take the mystery out of the a* and b* in CIELab.

There is a GATF test file that contains a Grey balance chart. They have several of these charts on the form. One is at the quarter tone. One is at a half tone. And one is at the three quarter tone. Each chart has a constant amount Cyan. Along one axis of the matrix there are increasing values of Yellow. Along the other axis of the matrix are increasing values of Magenta. First you must find one of these printed charts that was printed on a typical printing press that is close to Grey balance. The press sheet must also contain a solid Magenta, Yellow, and Cyan. Next you must find a spectrophotometer that can read out CIELab.

Follow these steps:

1. Measure the solid Magenta patch. Notice that the b* value is near zero. Notice that the a* value is around 80.

2. Measure the solid Yellow patch. Notice that the a* value is near zero. Notice that the b* value is around 90.

3. Measure the solid Cyan patch. Notice that both a* and b* are both negative.

4. Measure each patch alone the side of the Grey balance chart where Magenta is increasing. Notice that a* is changing and that b* is almost constant.

5. Measure each patch alone the side of the Grey balance chart where Yellow is increasing. Notice that b* is changing and than a* is almost constant.

6. Find the square in the matrix where both a* and b* are close to zero. This is the square that the "standard observer" would call Grey. You may see Grey as a different square. Ask other people to pick the Grey square. You may find that different people pick different squares clustered around the "standard observer". This is called viewer metamerism.

If you are trying to match two swatches of color and have the lab readings from the two swatches you can make corrections in the ink space by following this algorithm:

Assume that you have two patches. A master patch and a proof that you are trying to color correct.

If the a* of the master is greater than the a* of the proof, then you need to add Magenta to the proof.

If the b* of the master is greater than the b* of the proof, then you need to add Yellow to the proof.

If the a* and b* of the master are both equally less than the proof you need to add Cyan to the proof.

You can scale the amount of the move by measuring the full a* or b* range from the white point of the paper to the solid and scaling it to 100% in the ink space.

You can now measure in CIELab and correct in the CMYK ink space.

Now that you have finished this experiment find all of the text books that refer to the a* axis of the Lab space and have the positive end of the scale labeled RED and scratch out the word RED and replace it with Magenta. While not exact it is much closer to the truth for most printing inks.

The bottom line...a* is the Magenta-Green axis. Plus a* is Magenta. Minus a* is green. b* is the Yellow-Blue axis. Plus b* is yellow. Minus b* is blue. When a* and b* are both negative it is Cyan.

Hope this helps,

Ray

P.S. this whole experiment can be done in Photoshop by working in a CMYK space reading the Lab values in the information panel.

Creo

Ray Maxwell | Senior Color Systems Engineer, Inkjet Printing

4225 Kincaid Street | Phone (604) 451-2700 ext. 2004

Burnaby, B.C.

Canada V5G 4P5

IMAGINE CREATE BELIEVE

___________________________________________________________________________

Dan will surely weigh in to this one, but I have a little mnemonic:

b is for blue : up is yellow, hey, you're a mellow fellow; down is blue; when you have the blues

a is for amore (love): more is a hot magenta; less is a jealous green

And when I forget the mnemomic I just crank the curves in the channels way up and down, and it's obvious right away.

But L is the key, because if you don't have good shadows and good highlights you can work like L on the other channels and you'll still have an image from, well,limbo at best.

-- Gerald

___________________________________________________________________________

Thanks to all for the enlightenment on Lab mode. Based on what everyone has written here and offline, I'm beginning to believe that much of the confusion could be eliminated if Adobe or Binuscan (or your favorite Photoshop competitor) would replace the CMYK/RGB-centric curves dialog box with a "global color model."

Sure, it would be a bit graphically intensive. However, I believe it would make working in Lab easier.

--

Eric Curtis M. Basir (Bond)

Photo Grafix

http://www.abetterreality.net

(847) 673-7043

___________________________________________________________________________

Ray,

Can you tell us why some people define the a* as the red-green axis, instead of as the obviously logical magenta-green axis? I've wondered if it's just due to RGB-speak which digital artists and marketers tend to use; nobody ever *says* magenta when they're talking about a reddish hue. And the wrong information was passed on to others?

But ignoring the "PLUS blue" that's in magenta is a very bad idea because blue is such a big problem-color in digital photos.

And why isn't there a "red-cyan axis c*"? Redundance? I'd prefer redundance to mystery.

When I started inkjet-printing two years ago, the hype at that time was: Inkjet-printers are RGB-tools; you don't need to worry about CMYK any more. Of course, this belief..or hype..(false hope) had been enabled by the addition of printer driver PROFILES that were built around an understanding(?) of the 3D LAB color space. I believe the hype blinded most of us, or made us lazy.

I also believe that most digital push-botton manipulation-routines create a color-mess, and when color-management add-ons are used on top of that to average out the mess, color-quality is reduced.

Does anyone here know of any DOS, Windows, or Linux editing software that does nothing but LAB-editing?

Pat Thompson

___________________________________________________________________________

Pat Thompson writes,

The scientists have had it wrong for quite some time. Even my 1981 edition of Billmeyer and Saltzman states, "[The A channel] is a red-green opponent coordinate, for which positive values denote redness and negative values, greenness. Similarly, the [B channel] is a yellowness-blueness coordinate..."

AFAIK the source of the correction is me. In my book Professional Photoshop 5 (1998) I had a chapter all about LAB in which I repeated the red-green designation. But, as I worked more and more with LAB, I realized that this was a mistake, so in the revised edition of PP5 that was released in 1999, I changed it to magenta-green, and the rest of world, I think, followed suit.

I think that just two opponent-color channels is confusing enough. Adding a redundant channel adds a huge layer of complexity--just think about the impact of the black channel in CMYK. I've played around with an RFGB space (F stands for fleshtone color) with interesting results, but I don't think it's for the average user, who probably has enough trouble with FRGB, if you catch my drift.

Magenta-green is a much better choice than red-cyan, IMHO, because it really emulates the way the human visual system works. Last year, I did some experiments with color-blind people that eventually I'll write to the list about. But one of the things I was trying to find out was whether there's a similar term screwup with regard to people who have said visual disability. The scientists always call it "red-green" color-blindness, but in fact it's magenta-green. Some of the most seriously impaired people couldn't see a difference even when I inverted the A channel, but the minute I did anything to the B channel they caught it.

Editing in LAB is an amazingly powerful tool, but it isn't right for every image. So I don't know why you would want to restrict yourself to that space.

Dan Margulis

___________________________________________________________________________

--- Dan Margulis wrote:

Thanks for the information and history. And I wish the rest of the world HAD followed suit, but it hasn't. Now I'm wondering if the a* discrepancy is affecting present-day color-management systems and profiles that use LAB as a reference space. If the programmers use numbers, only, the difference between red and magenta probably makes no difference; but if the programmers don't use numbers...

No, I don't intend to limit myself to LAB. I just don't want to reinstall PS v6 only to use LAB...long story.

Pat Thompson

___________________________________________________________________________

Since I learned about printing from guys who were missing fingers, I always take the word "red" with a grain of salt. They may be talking about red or it's just as likely they're talking about "Process Red." Which is magenta. Your reference above was probably not due to this confusion but still the words red and blue do double-duty as both scientific certainties and aesthetic generalities. To be more precise when speaking and writing is a lofty goal. Good luck.

Off topic:

Am I off base wishing for the ability to adjust LAB curves without actually changing modes? Is it impossible for the curves box to offer 10 channels [RGB, CMYK, LAB] at once?

-Scott

___________________________________________________________________________

Scott Larsen wrote:

...

If Windows works for you, check out the curvemeister plugin, which comes close to doing this.

Mike Russell

www.curvemeister.com

www.geigy.2y.net

___________________________________________________________________________

There is always Photoshop's "Fade to Luminosity" feature.

___________________________________________________________________________

Or it's opposite - color blend mode.

OK in RGB, quirky in CMYK but still better than nothing, these seem to be on the fly simulations.

LAB _is_ really different in how the L is treated, for colour to grey moves L and LAB seems to offer a lot power for edits, which I will have to explore more...

Dan's latest MakeReady article in Electronic Publishing ep.pennnet.com provides some deeper commentary on the pros/cons of color/luminosity blends vs. LAB edits.

There is an old copy of Photoshop 3 Bible sitting on the shelf at work, I recently flipped through it and Deke had a suggestion for learning how the AB channels work...make a new blank file and generate flat tints and or gradations in the AB channels - as an aid to comprehension of what the tonal values mean by themself and how they interact with each other when combined.

As a beta tester on the curve plug product that Mike Russell authored,

I too would suggest having a look if you are PC based - there are

some nice features.

Stephen Marsh.

___________________________________________________________________________

"Stephen Marsh" wrote:

So what you're saying is that there's always more than one way to skin a cat? Got it. I'm absorbing--slowly--Dan's book and have fiddled with these techniques a little. I'll have to try harder to work them into my arsenal.

Noop. Mac. Thanks though.

-Scott Larsen

___________________________________________________________________________

Stephen writes,

This method is a start but it's likely to cause confusion. Remember, in LAB it's very easy to make colors that are not just out of the CMYK gamut but way out of the RGB gamut as well, meaning that the file is going to be calling for a color that's undisplayable, and yet Photoshop will try to display it.

If we use a blank L channel, any move we make in the A or B results in such a color, even in RGB. That is, 100L50A50B roughly describes the blinding red of a laser beam in real life. No monitor can display such a red: the brightest color that it can display is a pure white, with all its phosphors at maximum intensity. To get a red, it has to spike its green and blue guns and the overall feeling becomes much darker. So, we'd be calling for an utterly impossible color, and the results would be unpredictable at best.

Same way with gradients. Anything more than about halfway up or down the curve in A or B is likely to be out of gamut, even where the L channel is optimal for the particular color. Anyplace else, even fewer AB colors are in gamut.

I think the best way to understand how it works is to take a real picture, select a rectangular area, and watch what happens when you fool with the A and B channels.

Dan Margulis

___________________________________________________________________________

Stephen writes,

I found it quite clear and handy:

http://www.ledet.com/margulis/PSW_Dan_Magic_of_LAB.pdf

--

Claudio Corvino