Untitled Document

After reading Dan's LAB book three times, there's one thing I still just don't get. The book keeps talking about how, for example, "... as experienced LAB practitioners, we know that the key phrase is not 'more yellow than' but 'more yellow-as-opposed-to-blue than.'" (p.206).

What confuses me is that while LAB has four primaries (magenta, green, yellow & blue), a color apparently always is designated only by two primaries at a time; a color specification never involves three or four of the primaries. Only one of the opposing colors at once within each of the two LAB color channels (a & b) can be used to denote a color.

So what sense does it make to say that a color is "more yellow-as-opposed-to-blue"? (Or how does that non-trivially convey any more information than to say merely, "more yellow than"?) Any color having a value in the b channel is described as either yellow or blue, not a combination of the two, like one *can* have a combination across the a & b channels, e.g., some yellow with magenta or green, or some blue with magenta or green.

It does make sense to me to say that a LAB color is, e.g., more magenta-than-yellow or more green-than-blue, because both colors simultaneously can be involved in the color definition, with one having a higher value than the other. But any color involving, e.g., yellow cannot contain *any* blue whatsoever, right? Each opposing color in the two a & b channels is mutually exclusive from its sister color, no?

As a corollary question, is it the same in the actual human opposing-color visual subsystem upon which LAB is based? That is, does a color perceived containing, e.g., magenta, preclude any green component at the same time?

I'm sure I'm missing something basic here, and there's a simple reconciliation. Can someone straighten me out? Thanks!

George Machen

___________________________________________________________________________

Some of what you say could relate to hue angle, as in Lch, vs. thinking in straight a/b terms.

One example I can think is that if took a pure yellow of, say, a*0 b* +90 and then reduced it to b*+40, could I not describe that yellow as being "more blue" since I reduced/subtracted the b* value? Any subtraction of b* could be alternately described as "more blue" or "less yellow" regardless whether the b* value was in fact a negative value. Could be splitting a/b hairs here, but I don't see an issue with describing a reduction in b* as being more blue whether or it actually has a negative b* value in absolute terms.

Of course, the "Lch" in me (way better at describing what's really going in in L*a*b* in my opinion) says I've simply reduced the Chroma of the yellow, nothing more!

:-)

Regards,

Terry Wyse

_____________________________

WyseConsul

704.843.0858

http://www.wyseconsul.com

http://www.colormanagementgroup.com

___________________________________________________________________________

Hello George Machen,

I think what Dan means in p.206 is that the left-side trees and the red logo although looks green and red in fact have a higher b value than the least colorful parts of the train which seems to be "more yellow than", say, the green tree or the red logo and definitely the blue. So the left-side tree and logo are in fact "more yellow as opposed to blue". Saying the least colorful part "more yellow than" may be misleading. Would Dan please correct me if I'm wrong !

You may find the answer to this question in P.24-25 of Bruce Fraser's book "Real World Color Management (2nd Edition)" under the topic "opponency and trichromacy", especially Fig. 1-14 which says "signals from cones either excite or inhibit second layer of neurons, producing opponent signals".

Kind regards,

Wai-hong Chung from Hong Kong

___________________________________________________________________________

Have you tried it? If you try it (shift the "b" curve in a manner that decreases yellow, you will see that the overall hue of the image becomes more bluish. I don't see that there is much more to this issue than that.

Mark Segal

___________________________________________________________________________

George,

If, as you say, you're simply reducing the saturation of yellow my question would be what's wrong with looking at it that way ??? In other words the yellow hue and its saturation without any reference to its lightness or darkness.

I like what Terry wrote, "Of course, the "LCH" in me (way better at describing what's really going in L*a*b* in my opinion) says I've simply reduced the Chroma of the yellow, nothing more!

Lee Clawson

2/\V/\7 Studio

___________________________________________________________________________

From the point of view of where I'm coming from (possibly still confused), reducing the yellow value, well, just reduces the yellow saturation, doesn't increase the blue value. In my mind the latter would create a green (were in LAB both yellow & blue components allowed at the same time in a single color designation).

Describing it as more blue seems to me would have to mean allowing all four LAB primaries to define a color, not being confined to just one each from the two color channels.

Now, in RGB one can reduce the yellow by actually adding more blue, but then luminosity is involved, too.

I know I still must be missing something obvious.

George Machen

___________________________________________________________________________

Hello George Machen,

I think what Dan means in p.206 is that the left-side trees and the red logo although looks green and red in fact have a higher b value than the least colorful parts of the train which seems to be "more yellow than", say, the green tree or the red logo and definitely the blue. So the left-side tree and logo are in fact "more yellow as opposed to blue". Saying the least colorful part "more yellow than" may be misleading. Would Dan please correct me if I'm wrong !

You may find the answer to this question in P.24-25 of Bruce Fraser's book "Real World Color Management (2nd Edition)" under the topic "opponency and trichromacy", especially Fig. 1-14 which says "signals from cones either excite or inhibit second layer of neurons, producing opponent signals".

Kind regards,

Wai-hong Chung from Hong Kong

___________________________________________________________________________

Have you tried it? If you try it (shift the "b" curve in a manner that decreases yellow, you will see that the overall hue of the image becomes more bluish. I don't see that there is much more to this issue than that.

George Machen wrote:

___________________________________________________________________________

George,

If, as you say, you're simply reducing the saturation of yellow my question would be what's wrong with looking at it that way ??? In other words the yellow hue and its saturation without any reference to its lightness or darkness.

I like what Terry wrote, "Of course, the "LCH" in me (way better at describing what's really going in L*a*b* in my opinion) says I've simply reduced the Chroma of the yellow, nothing more!

Lee Clawson

2/\V/\7 Studio

___________________________________________________________________________

OK, I tried it, a simple move in curves, and the image indeed went blue, just as would happen in RGB when reducing the yellow.

But then I noticed in the info palette that the newly blue areas had gone negative; blue explicitly had been introduced. And then I realized, as Dan took such pains to warn in the LAB book, curves in LAB are different! *Two* primaries are being controlled in each color channel, divided at the zero value.

So then I pinned the b channel curve below zero such that moves only could reduce yellow and not simultaneously add blue. This time those brown shoes on the dark yellow floor didn't turn blue; rather, the shoes went magenta and the floor went more neutral, just as I would expect, in retrospect.

So I conclude that reducing yellow in LAB is not the same as doing so in RGB: in and of itself it only desaturates; it doesn't go blue. Is my little test conclusive & fair, or am I still misunderstanding things?

Thanks!

George Machen

___________________________________________________________________________

You may be confused because LAB doesn't use primaries to describe color at all. In RGB or CMYK, we look at color in terms of HOW MUCH of each primary color is used to make up a particular hue. In LAB, we look at color only in terms of the BIAS of each the two components, not four primaries.

One component measures the blue vs. yellow bias and the other measures the magenta vs. green bias. Since luminosity has been taken as a separate function, we don't concern ourselves with how much of a color is present in the mix, only the color bias is important, and it only takes two components to do that. In LAB, all yellows except for the most theoretically pure yellow are also blue, etc.

The question I have for the color scientists among us is why only these particular axes were chosen. Scientists like to reduce equations to their simplest forms, and while the simplicity of LAB may be questionable to some even as it is, I imagine there are situations where a red/cyan and orange/purple might be convenient to have as c and d channels. Anyone like to have an LCD or LABCD color space?

John Castronovo

___________________________________________________________________________

George Machen writes,

I would say that you have achieved the right result but still have not found the proper state of mind. The other people who have responded are correct, but let me try some different words.

The LAB primaries are not inks. They do not mix with one another to create other colors. They are chosen because they are direct opponents of one another. If you move away from one, you move toward the other. So, your phrase "it only desaturates; it doesn't go blue" is not correct--it *does* go blue, because if you are moving away from yellow, so you are moving toward its opponent. A desaturated yellow is more blue than a saturated one, in the LAB construction.

It might be helpful to look at it this way: a brilliant blue is the least yellow thing you can possibly imagine. So, instead of calling it blue, call it "anti-yellow". Now, in your B channel, positive values are yellow; negatives are anti-yellow, zero is neither one.

Desaturating a yellow, in LAB, means that you have added anti-yellow to it. So, by locking down the center of the B curve as you did, you added anti-yellow to the bottom half of the channel, but not to the top half. So things that were originally yellow became more anti-yellow (more blue; less saturated), and things that were originally anti-yellow remained unchanged.

Dan Margulis

___________________________________________________________________________

All this talk about opponency color bias and all yellows containing blue except for the purest yellows suggests that the b channel component of a color is a *mixture* of yellow and blue. But it seems to me that this cannot be right. A positive b channel is of solely yellow hue, losing saturation with declining positive value until it goes negative, whereupon its hue henceforth is solely blue with no yellow mixed in. Didn't my little curves test with the shoes demonstrate this? Aren't the a & b channels just a very clever but arbitrary construct engineered to put two LAB "primaries" instead of one into each channel (but the two still are isolated, bordered at zero!), and to make convenient the representation & manipulation of curves and overlay blends?

I still can't see how greater information is being conveyed by saying "more yellow-as-opposed-to-blue than" instead of by merely saying "more yellow than." When the b channel is positive, no blue is involved whatsoever. Or at least this is what I can't get past.

Another example: When I exclude the top layer with the b channel Blend-if slider, I exclude yellow and only yellow until the slider ventures into the blue side.

George Machen

___________________________________________________________________________

On 10/7/06, George Machen wrote:

Right. Yellow on one side and blue on another. Negative is blue, positive is yellow, zero is neutral. I like to remember it by thinking that the Kansas City Chiefs are positive and the Seattle Seahawks are negative.

Why not? LAB is a wierd beast, unlike any other colorspace. It also makes the least sense, so you just have to go with it.

Now you're mixing LCH with LAB. The idea is right, but sorry to say Blue opposes Yellow in human vision.

NO.

I wouldn't say arbitrary at all. It's not like Magenta/Green and Blue/Yellow were picked randomly.

Now you're getting down to writing style. Dan likes to be specific like that.

NO.

Try. :-)

LAB is a great correction space, but you have to be willing to accept that it doesn't act like anything else you've seen. Just stick with it, try not to get bogged down in little details, and you'll do fine.

-----

J Walton

___________________________________________________________________________

I took it that LAB was setup in this particular manner because if something is blue it can't be yellow and if something is red it can't be green. kind of how paint mixing gives you different colors, but in this case equal amounts of green and red make white, and equal amounts of blue and yellow make white. The difference in colors having to do with additive versus subtractive color spaces.

Matthew Rigdon

___________________________________________________________________________

George,

Honestly I don't think I can help make this clear. In fact after reading your post a second time I wonder if I don't feel confused like you about greater information is being conveyed.

So even with impending shared confusion I think back to a question of saturation. Lightness/darkness ("L" channel) is not included and we know the hue (yellow). I ask myself what will be added to reduce the saturation. Asked a different way, "what's the gray component of the color.

Lee Clawson

2/\V/\7 Studio

___________________________________________________________________________

Eureka! To the contrary, Lee, what you said precisely is what I needed to make it clear! Indeed, clarity descended like a rock and hit me in the head just as quickly.

Your putting it in terms of what it takes to reduce a color's saturation - what is the "gray component" of a color - is the key to grasping LAB opponency. Blue kills yellow, green kills magenta ... (cyan kills red, for that matter).

Thus, reducing the yellow component of a LAB color (desaturating yellow) is functionally equivalent to adding blue, in that sense. The confusion arises only because the explicit blue hue becomes visible only after the decreasing yellow saturation passes through zero and the b channel goes negative.

In other words, with a LAB color, increasingly adding its opponent color manifests only in decreased saturation until it goes completely neutral; only thereafter does the opponent color's actual hue appear.

(That also reconciles my little curves test with the brown shoes on a yellow floor, where I pinned the blue.)

By Jove, I think I've got it!

George Machen

___________________________________________________________________________

On Saturday 07 October 2006 13:39, john castronovo wrote:

I was curious about this, too. Why *does* LAB use green/magenta and yellow/blue? Could other axes be used?

Bjorn Helgaas

___________________________________________________________________________

Bjorn writes,

It would be possible to substitute a red-cyan axis for either the A or the B. Adding more axes without deleting the existing ones might make color correction easier, however it would defeat the purpose of using LAB as a reference space, because it would then be possible to create the same color in multiple ways.

The use of green/magenta and yellow/blue is traditional because these two appear to have better correspondence to human visual response than red-cyan does. Therefore, you find green/magenta and yellow/blue in most reference spaces, such as LUV, as well as in each successive iteration of LAB.

There is also the practical reason that whichever two of the possible three axes are chosen allows more room for the colors of the omitted axis than those of the chosen two. So, there are far more possible LAB combinations that make red than that make yellow. We have to reproduce reds more frequently than any other color, greens second, blues third, followed bv YMC in that order. Greens and blues, however, fall in rather a narrow range: natural greens are always yellowish, never cyanish and rarely pure green.

In comparison, we commonly have to reproduce not just pure reds, but magentaish reds (roses, for example) and yellowish reds (human faces). So we can really use the extra space for creating reds that the current LAB structure provides.

Dan Margulis