Photoshop Lab Color: Ch. 2

ruttrutt Registered Users Posts: 6,511 Major grins
edited October 25, 2008 in Finishing School
[size=+4]Don't Panic![/size]

Chapter 2 is a technical explanation of LAB and its relation to the more familiar RGB and CMYK colorspaces. I think the most important thing to keep in mind when you read this chapter is that Dan thinks the actual numbers are more important than you probably need to. Ideally, you'll come away from reading this chapter with a basic understanding of what each of the channels of a LAB image mean. Ideally you will also have an idea of how LAB is structured. (I think this is very fun to get, since it is based on an incredibly elegant idea.) There are a few things about LAB numbers which are extremely useful and also easy to learn. But don't get hung up if you can't look at a color and tell it's LAB values or visa versa. You don't need this in order to use LAB effectively (you learned this when you learned the recipe in Ch. 1). To the extent that this skill is helpful to you, you will learn it over time.

I'm going to give my version of what I think is important and/or interesting about LAB. I'm not as good a writer as Dan is, and I don't have his depth of understanding on this topic, and I haven't taught as many people as he has. But maybe I can provide a less intimidating explanation for those who find Dan too technical. (But then maybe I won't succeed.) In any case, I'll give it a try. One thing I can certainly do is to try to answer questions.

LAB is a colorspace

A colorspace is a system which uniquely specifies colors as a list of numbers. For example, RGB uses three values between 0 and 255, one each to specify a brightness of red, green, and blue light in the color. the list R=255, G=255, B=255 represents the lightest white and R=0, G=0, B=0 indicates the darkenst black. In RGB we know a color is neutral if it is equally light in all three channels, that is if R = G = B. Then it is not more red than blue and not more green than blue or red.

LAB is quite different from RGB or CMYK in that it has a separate channel ("L" for Luminosity) for brightness. So in LAB brightness is completely independent of the color component. The values of the L component of a color run from 0 for the darkest possible to 100 for the lightest possible.

Having a separate luminosity channel is wonderfully useful. It means that we can sharpen without creating color artifacts, gives us a brightness curve for controlling contrast independent of color, makes the shadow/highlight adjustment work when there is only data from a single channel, and lots more. It's also pretty intuitive.

The A and B channels are a whole different animal and less intuitive at first. The A channel goes from the greenest possible green (A = -128) to the most magenta possible magenta (A = 127). In between, it passes though A = 0 where there is exactly no green nor magenta. The B channel goes from the bluest blue (B = -128) to the yellowest yellow (B = 127). When B = 0, the color has no blue and also no yellow.

(Actually Dan writes (127) to mean -127 as an accountant would. He says this make it easy to read.

The most easiest and most useful thing to know about LAB numbers is that A = 0, B = 0 implies that a color is neutral, either white, black, or some shade of gray depending on it's L value. Set up the A and B curves so that some point on your image has A = 0, B = 0, and you can control it's brightness with the L curve and the point will stay neutral.

Another thing to learn at this point is that in both A and B, warmer colors are have positive numbers (and are shown as the "light" sides of the curves) and the cooler colors have negative numbers (and are shown at the "dark" sides of the curve.) Yellow and magenta are the warm colors; together they make red. If both the A and B numbers are positive, the color will be warm, some reddish, yellowish thing. Blue and green are the cool colors and have negative values.

That's really all you have to know about this chapter in order to keep reading and understanding the book. I suggest at least skimming it and read on if you aren't having fun. Don't worry, you can always reread it if you need to (probably you won't.)

Oh, two more things that you might want to learn at this point (although it will come up again in a more friendly context in Chapter 3):
Healthy fleshtones are always some shade of red, and so they will have A and B both positive. The greens of vegatation are yellowish green (actually) and so will have negative A and positive B, with the B not more than the A.

The rest of what I have to say is really of intellectual interest only. People who have no patience for abstract theoretical knowledge may safely stop reading at any time without missing anything important.

Opponent Colors

The A and B channels of LAB are opponent color channels. A runs from green to magenta which are opponent colors, which means that green and magenta cannot both be present in the same color. There cannot be any such thing as a greenish magenta. Shine a perfectly green light on a perfectly magenta spot of color and you will see black. Magenta pigment doesn't reflect green light.

Since a color cannot be both green and magenta (or blue and yellow) at the same time, LAB only has to specify how intensely green or magenta the color is. More negative values specify more intense green, 0 means no green nor magenta, and more positive values specify more magenta. Simalarly for the B channel, where more negative numbers specify more intense blue and more positive number specify more intense yellow.

Why no C channel?

When I first learned about LAB, I wondered why there was no channel for the third pair of opponent colors, cyan and red. I imagined that a C channel could run from cyan to red in the same way the A and B channels work with their opponent colors. I even asked Dan this question in person once. His answer was, that it's much more useful the way it is.

Now that I understand things a little better, I know that answer was just used to silence a question that Dan didn't want to waste time on. The real answer is that a C channel is redundent. Given L, A, and B values, a color is completely specified. It will be red if A and B are both positive, cyan if they are both negative. And the brightness is given with the L channel.

It would be possible to have an ABC colorspace, I suppose. with A=127, B=127, C=127, we'd have the brightest most vivid red possible. With A=127, B=127, and C<=0, we'd have black. (Oh well, I doubt this makes sense to anyone but me. Let me know if it does make sense to you and we can get drunk together sometime and talk about it.)

We don't have ABC because it is so very much more useful to have an L channel than it ever could be to have a C channel.

Impossible Colors

Anybody still with me? OK, if you made it this far, you probably are a real nerd and boy are we going to have a little fun right now.

One bizarre thing about LAB is that it's gamut is impossibly wide. Literally. There are several thought experiements which show this. Consider the color L=100, A=0, B=-128. This is a color as bright as possible and also as blue as possible. Could such a color exist? Suppose it did and suppose we had a light just that color. What would happen to it if we added red and green light? Wouldn't it get brighter? But it was specified to be as bright as possible...

Or take another example, one which Dan actually gives in this book (the previous example was from a post of his to his colortheory mail group.) Consider a color as yellow as possible but pretty dark. What does this look like? There really is no such thing as a dark yellow. It looks brown. (Remember the tennis ball color theory thread from a little more than a year ago?) But you can specify it in LAB.

The gamut of LAB vs other colorspaces

LAB's gamut is by far the widest of the photoshop colorspaces. As we just saw, it's easy to express colors in LAB which have no expression in RGB let alone CMYK. What happens to these colors as we move an image to other colorspaces? Dan spends most of the second half of the chapter on this, but in short, I think the answer is that moving between LAB and RGB won't influence how an image looks on your monitor. The impossible colors go away, but they weren't displayed in the first place.

The fact that you can go so far out of gamut in LAB can actually be useful. Dan gives a cool example with a sunset. The sun itself in such pictures is really well out of the gamut of any print where the brightest possible color is the paper with no ink. And no monitor is as bright as the sun. And the sun is also yellow. By using a very yellow and bright color for the sun (applied via very steep curves) he gets a much better looking sunset than would be possible with an RGB correction.

Conclusion
[size=+4]Don't Panic![/size]
I know I just wrote a lot of words and no pictures and that Dan quizzed you on what numbers represented what colors and acted as if you needed to know or your were a waste of precious minerals and water. But really there isn't that much in this chapter you need to know right now. The structure of the book demanded a technical explanation of LAB at this point, and Dan did his best to provide it. He thinks about colors with numbers, but you don't have to in order to benefit from his knowledge.
If not now, when?
«1

Comments

  • AndyAndy Registered Users Posts: 50,016 Major grins
    edited September 13, 2005
    very nice work, rutt. you're making it very easy to understand. this reading group, these virtual "classes" are invaluable. thanks for putting out the time to do this, i, and i'm sure everyone, really appreciates it! thumb.gifthumb.gif
  • pathfinderpathfinder Super Moderators Posts: 14,694 moderator
    edited September 13, 2005
    "I know I just wrote a lot of words and no pictures and that Dan quizzed you on what numbers represented what colors and acted as if you needed to know or your were a waste of precious minerals and water. But really there isn't that much in this chapter you need to know right now. The structure of the book demanded a technical explanation of LAB at this point, and Dan did his best to provide it. He thinks about colors with numbers, but you don't have to in order to benefit from his knowledge."

    Thanks John for your excellent description of Chapter 2. I feel SO much better now - I WAS feeling like a waste of precious minerals and water:D

    I understand what you have written quite clearly about the structure of LAB, but I was finding it somewhat difficult to immediately visualize L 80 A (100) B 27 - but I suspect Dan looks at those numbers all day long and they begin to become second nature after enough time. But I can't do that yet and I'll bet I'm not alone. :):

    I am looking forward to more descriptions of impossible colors - very dark yellow, very brite royal blue, etc.
    Pathfinder - www.pathfinder.smugmug.com

    Moderator of the Technique Forum and Finishing School on Dgrin
  • ginger_55ginger_55 Registered Users Posts: 8,416 Major grins
    edited September 13, 2005
    I feel SO much better now - I WAS feeling like a waste of precious minerals and watericon10.gif
    ___________________________

    PF, that is VERY funny!

    I second/third the thanks you are getting, Rutt, especially after reading Pathfinder's description of numbers.

    ginger
    After all is said and done, it is the sweet tea.
  • SamSam Registered Users Posts: 7,419 Major grins
    edited September 17, 2005
    Rutt,


    Thank you for taking the time to write this explanation. I will reread this at a later time. I would like to get to the point were I had an idea of what the color is by looking at the numbers. Not there yet.

    Sam
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 19, 2005
    I am skeptical about the necessity to perform these adjustments in Lab for one primary reason -- that being that photoshop thinks in Lab whether you're working in it or not. The tonal adjustments that you make in the L channel are no different than those you make to the RGB composite (and in fact any filter or process that you fade or blend to luminosity specifically applies only to luminance, i.e. the L channel). The size of the color space is a big issue, but if you shoot in Adobe RGB and open your RAW image in Prophoto RGB you have a color gamut comparable in size to Lab.

    So I'm not suggesting that Lab is any worse, except by virtue of the fact that it's less intuitive and many of the filters won't work in it. Also, for making color adjustments, you may have more control if you are able to modulate 3 channels simultaneously (R, G, and B) rather than two (a and b).
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Plasmodium wrote:
    I am skeptical about the necessity to perform these adjustments in Lab for one primary reason -- that being that photoshop thinks in Lab whether you're working in it or not. The tonal adjustments that you make in the L channel are no different than those you make to the RGB composite (and in fact any filter or process that you fade or blend to luminosity specifically applies only to luminance, i.e. the L channel). The size of the color space is a big issue, but if you shoot in Adobe RGB and open your RAW image in Prophoto RGB you have a color gamut comparable in size to Lab.

    So I'm not suggesting that Lab is any worse, except by virtue of the fact that it's less intuitive and many of the filters won't work in it. Also, for making color adjustments, you may have more control if you are able to modulate 3 channels simultaneously (R, G, and B) rather than two (a and b).


    RGB is three color values. The a and b curves are four. A is Magenta and Green, B is Blue and Yellow.

    LAB is actually more intuitive, once you dig in and learn it. It's not a replacement for RGB, but a super-charged addition that can dramatically improve the quality of your shots faster and often better than doing the work in RGB.

    And the adjustments you make in L ARE VERY DIFFERENT than the ones you make in the RGB composite. RGB effects color and adds artifacts, LAB doesn't.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 19, 2005
    DavidTO wrote:
    RGB is three color values. The a and b curves are four. A is Magenta and Green, B is Blue and Yellow.
    That's sort of semantic. The composite of Lab, and the composite of RGB constitute the entirety of the colors in the image. With RGB you can independently modulate 3 color variables, whose composite produces the final color. Lab only lets you independently modulate 2. Now I don't have the same intimidation about Lab that some photoshop books express (like in my book by Deke McLelland), but at the same time I have found I have more flexibility when working in RGB.

    The other thing to be aware of (though it's easy to get around) is that when you convert Lab to RGB, even if you were working in a big RGB color space like Adobe or ProPhoto, is that you automatically get plunked in sRGB by Photoshop (unless you do Edit>convert to profile and convert yourself into a big color space like Adobe RGB).
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Plasmodium wrote:
    That's sort of semantic. The composite of Lab, and the composite of RGB constitute the entirety of the colors in the image. With RGB you can independently modulate 3 color variables, whose composite produces the final color. Lab only lets you independently modulate 2.

    That's false. LAB lets you independetly modulate 4.
    Now I don't have the same intimidation about Lab that some photoshop books express (like in my book by Deke McLelland), but at the same time I have found I have more flexibility when working in RGB.

    Cool. RGB is mighty powerful, too.
    The other thing to be aware of (though it's easy to get around) is that when you convert Lab to RGB, even if you were working in a big RGB color space like Adobe or ProPhoto, is that you automatically get plunked in sRGB by Photoshop (unless you do Edit>convert to profile and convert yourself into a big color space like Adobe RGB).

    Understood. The purpose of LAB is not to produce out of gamut colors (although you have to be careful about that), but to make your images dramatically better quicker than you can with any other method.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 19, 2005
    DavidTO wrote:
    That's false. LAB lets you independetly modulate 4.
    I indeed stand open to correction, but this is how I conceive of it -- in, for instance, the b channel, if you ctrl-click on part of your image, producing a point on the curve, by adjusting that point you are simultaneously adjusting blue and yellow, not independently. They are on an axis, and moving away from one end necessarily moves you towards the other. I guess the exception to this would be when you make adjustments to the actual ends of the curve, rather than in the middle, because by definition you're eliminating one of the ends of the axis (though you do change the slope of the curve and thereby all the blue-yellow pixel relationships therein).
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Plasmodium wrote:
    I indeed stand open to correction, but this is how I conceive of it -- in, for instance, the b channel, if you ctrl-click on part of your image, producing a point on the curve, by adjusting that point you are simultaneously adjusting blue and yellow, not independently. They are on an axis, and moving away from one end necessarily moves you towards the other. I guess the exception to this would be when you make adjustments to the actual ends of the curve, rather than in the middle, because by definition you're eliminating one of the ends of the axis (though you do change the slope of the curve and thereby all the blue-yellow pixel relationships therein).


    One end of A is magenta, the other green. The center is neutral. So changing one end of the curve effects that one color. There can never be magenta and green in the same place. It's one or the other.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 19, 2005
    The middle of the curve is neutral only when the input and output values are equal. If you move the dead middle of the a curve upward from the neutral point it takes on a magenta cast, and if you move it downward it takes on a green cast, and this occurs independently of moving the ends of the curve.
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • NikolaiNikolai Registered Users Posts: 19,035 Major grins
    edited September 19, 2005
    With all due respect
    DavidTO wrote:
    LAB lets you independently modulate 4.
    David, your are not correct, at least from a mathematical stand point.

    RGB, LAB and HSL are all 3-dimensional spaces. They are not identical (color gamut is quite different), but "structure" wise they are all 3D. The only reason CMYK adds the forth dimension is a practical impossibility to create a nicely looking black/gray print from the 3 colors (you know better than me that it creates so called "dirty black"), so they added another ink, thus making all the calculations in CMYK extremely cumbersome..

    However, LAB is 3D, not 4 (or 5, if you count L:-).Both Magenta-Green and Blue-Yellow are just one number each (BTW, thanks for the mnemonics tip on mAgentA and Bluethumb.gif Before I had a hard time to remember them, now it's a snap!). You can't change Blue "impact" without changing Yellow one for exactly same amount.
    Yes, those four colors do help you visualize this otherwise pretty abstract color space, but in fact they cover two dimensions only, leaving third to L..

    HTH
    "May the f/stop be with you!"
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Nikolai wrote:
    David, your are not correct, at least from a mathematical stand point.

    RGB, LAB and HSL are all 3-dimensional spaces. They are not identical (color gamut is quite different), but "structure" wise they are all 3D. The only reason CMYK adds the forth dimension is a practical impossibility to create a nicely looking black/gray print from the 3 colors (you know better than me that it creates so called "dirty black"), so they added another ink, thus making all the calculations in CMYK extremely cumbersome..

    However, LAB is 3D, not 4 (or 5, if you count L:-).Both Magenta-Green and Blue-Yellow are just one number each (BTW, thanks for the mnemonics tip on mAgentA and Bluethumb.gif Before I had a hard time to remember them, now it's a snap!). You can't change Blue "impact" without changing Yellow one for exactly same amount.
    Yes, those four colors do help you visualize this otherwise pretty abstract color space, but in fact they cover two dimensions only, leaving third to L..

    HTH

    You can correct magenta without effecting the green. Read up in Margulis, and I think you'll find that this is correct. It's Chapter 4 stuff.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • dandilldandill Registered Users Posts: 102 Major grins
    edited September 19, 2005
    rutt wrote:
    What about this: ...

    I added a step to the recipe: Shadow/highlight before converting to LAB. I find that doing this is often very helpful as much for highlights as shadows and it often makes writing that L curve a lot easier.
    That is a nice change. I'll give it a try. Thanks.
    Dan Dill

    "It is a magical time. I am reluctant to leave. Yet the shooting becomes more difficult, the path back grows black as it is without this last light. I don't do it anymore unless my husband is with me, as I am still afraid of the dark, smile.

    This was truly last light, my legs were tired, my husband could no longer read and was anxious to leave, but the magic and I, we lingered........"
    Ginger Jones
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Nikolai wrote:
    David, your are not correct, at least from a mathematical stand point.

    RGB, LAB and HSL are all 3-dimensional spaces. They are not identical (color gamut is quite different), but "structure" wise they are all 3D. The only reason CMYK adds the forth dimension is a practical impossibility to create a nicely looking black/gray print from the 3 colors (you know better than me that it creates so called "dirty black"), so they added another ink, thus making all the calculations in CMYK extremely cumbersome..

    However, LAB is 3D, not 4 (or 5, if you count L:-).Both Magenta-Green and Blue-Yellow are just one number each (BTW, thanks for the mnemonics tip on mAgentA and Bluethumb.gif Before I had a hard time to remember them, now it's a snap!). You can't change Blue "impact" without changing Yellow one for exactly same amount.
    Yes, those four colors do help you visualize this otherwise pretty abstract color space, but in fact they cover two dimensions only, leaving third to L..

    HTH

    I never said it was a 4D space, because I just do not understand the concept of a dimensional space in terms of color. I think the big difference is that the R channel is only red. Blue is blue and green is green.

    A is the continuum from magenta to green, passing through neutral on the way. It's one dimension of the colorspace (I guess, not really understanding this whole dimension thing), but it goes from one opposing color to the other. The range it covers is much larger than R,G or B or C,M,Y or K for that matter, because it describes the two opposing colors. And I have moved the magenta part of the curve with no effect on the green. It's awesome, and one of the best things about LAB.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • NikolaiNikolai Registered Users Posts: 19,035 Major grins
    edited September 19, 2005
    David,
    We're probably talking about different aspects.ne_nau.gif

    I think you are talking about a possibility to make a whole picture, or part of it, look more magenta while leaving the same impression of its green part.

    I was specifically referring to the mathematical aspect of dimensions. In essence, I was talking about an individual pixel, which in our case, represents a point in L-A-B space. Margulis or not, you can't make that *one* pixel become more ""magenta while keeping the same "green" value in it. You can keep some other pixels green, thus making an impression that part of the picture became more magenta while keeping its green value.
    But on a single pixel level - you can't. You only have *one* number for each of the channels.

    Once again, I'm not talking about the whole image editing. Layers, masks, and selections can do wonders, making some pixels to change their values while keeping other sets of pixels intact or "painting" them in a totally different way.

    But when you speak of "dimensions" and "spaces" - sorry, there are only 3 independent dimensions, and each pixel has exactly 3 coordinates in this wonderful space: L, A and B.

    Of course, each time we omitting the fact that in addition to COLOR we also deal with the SURFACE, since different pixels, while being the same or different color, are primarily distinguished by their spatial coordinates in 2D space of image plane. So, technically speaking, we *are* operating in 5D space (6D in CMYK case):):

    I hope I'm not getting in your way of explaining how to operate in LAB. You're doing a great thing with these tutorials. I only wanted you to be accurate even on the lowest, single pixel, level.

    Cheers!1drink.gif
    "May the f/stop be with you!"
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Nikolai wrote:
    you can't make that *one* pixel become more ""magenta while keeping the same "green" value in it.


    You can, and in fact, that's your only choice. If a pixel's magenta it is by definition not green. There's no green where there is magenta. So by definition, if you're changing the magenta values the green values stay the same, since you are not effecting any pixels that have green in them.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • NikolaiNikolai Registered Users Posts: 19,035 Major grins
    edited September 19, 2005
    I am afraid I can't agree with you..
    DavidTO wrote:
    You can, and in fact, that's your only choice. If a pixel's magenta it is by definition not green. There's no green where there is magenta. So by definition, if you're changing the magenta values the green values stay the same, since you are not effecting any pixels that have green in them.
    Let's be specific, shall we?
    We're talking about 1 (one) pixel, located at spatial coordinates (0,0) of the image (Lef-top corner for most of the computer-based graphical systems).

    Let's also talk about only A channel (actually, channels ARE dimensions, so since we have THREE channels we do have THREE dimensions, neh? Oh well, let's go on).

    This pixel in A is represented by ONE value only. It's a "signed" one, which means it can take both positive and negative values (unlike, say, L channel, which can take only positive values).

    System "interprets" positive values as green, negative ones a s magenta. Zero is neutral (gray).

    If this pixel had, say, value "+64", it means, with all other things being equal, it was on a green side. If you change it to "+32" you'll make it still green, but less green. Kinda less saturated green. Push it to zero, and it becomes gray. However, if you make its value "-32" it will be green no more, as it will be now interpreted as some magenta.

    I want to emphasize this once again - we're talking about a single pixel.

    In your reasoning, you're saying about "other pixels"
    you are not effecting any pixels that have green
    The others don't count. When you talking dimensions, you talking one point. One pixel. It has only 3 color coordinates. 3 channels.

    Think of it as of *one* bank account. You either have positive balance , or negative one (or zero in between). Adding money to this account gets you out of debt. If you owed $100 and you added $200 - there is no way you will still owe anything, on that account that is.
    There can be plenty of others, one person can have both huge debt and huge saldo.
    But that's for multiple accounts.
    For any single one - it's you you owe money, or the other party owes you (or you both are square:-). Can't have both.

    Did you see the new "associations" thread? mwink.gif
    "May the f/stop be with you!"
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 19, 2005
    DavidTO wrote:
    I never said it was a 4D space, because I just do not understand the concept of a dimensional space in terms of color. I think the big difference is that the R channel is only red. Blue is blue and green is green.

    A is the continuum from magenta to green, passing through neutral on the way. It's one dimension of the colorspace (I guess, not really understanding this whole dimension thing), but it goes from one opposing color to the other. The range it covers is much larger than R,G or B or C,M,Y or K for that matter, because it describes the two opposing colors. And I have moved the magenta part of the curve with no effect on the green. It's awesome, and one of the best things about LAB.
    They aren't just red, blue, green -- they are a continuum of more versus less green, more or less blue, and more or less red, which, are not only independently manipulated, but add up to a final pixel color.

    In lab, it's not more green independent of less magenta -- it's more green therefore less magenta. It's less yellow therefore more blue. At any individual pixel point that you select on the curve, you are making the final color of that pixel more blue by decreasing its yellow, and vice versa. You can prove it to yourself quite easily -- just take an image that is one solid color; better yet a neutral gray. Start a new document, then take the paintbucket and fill the image with neutral gray. Then look at what happens uniformly to the color if you adjust the a channel. If you move it one way it becomes more magenta, the other way more green. Those are the results of modulating a single variable, not two variables.
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Nikolai wrote:
    In your reasoning, you're saying about "other pixels"
    The others don't count. When you talking dimensions, you talking one point. One pixel. It has only 3 color coordinates. 3 channels.


    Because any single pixel can be only magenta or green. So at any location where it is magenta you can change the saturation of that pixel and not effect any green whatsoever. The curve is not a single color, but two, the opposites of magenta and green in this case. There is only magenta and by definition no green. Any green in the image WILL NOT BE AT THAT PIXEL.

    I never said that you could simultaneously change two colors at one pixel. Lab is designed so that the two colors on your curve are mutually exclusive at the pixel level.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Plasmodium wrote:
    In lab, it's not more green independent of less magenta -- it's more green therefore less magenta. It's less yellow therefore more blue.


    Wrong. The two colors magenta and green cannot exist at the same pixel, this is what gives you the flexibility.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 19, 2005
    DavidTO wrote:
    Wrong. The two colors magenta and green cannot exist at the same pixel, this is what gives you the flexibility.
    I will acknowledge that there is a neutral point beyond which it appears green on one side and magenta on the other.

    But that is irrelevant for two critical reasons.

    1) You are talking about complex colors that are determined by the other color channel and the lightness channel, not merely its green or magenta appearance

    2) most importantly, there is a continuum of values with a midpoint -- that midpoint is neutral solely because it's midway between all green and all magenta. But move it up a touch and it's that much closer to one end and reciprocally that much farther from the other. If you look at it in isolation, you can convince yourself that magenta and green are being manipulated independently, but it's really not true when you consider that this is a whole spectrum being added to the totality of a composite pixel.

    Better yet, think of it this way. By your rationale can you manipulate green and magenta at one point at the same time? If you're on the green side of the neutral midpoint, by your logic, you cannot manipulate magenta at all! (again, talking about one pixel, not the entire image).
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 19, 2005
    Plasmodium wrote:
    If you're on the green side of the neutral midpoint, by your logic, you cannot manipulate magenta at all! (again, talking about one pixel, not the entire image).


    That's exactly what I'm saying. If you want to manipulate the magenta, you need to be on the magenta end of the curve.

    If you move the curve as a line, then you'll effect both, since you are moving both ends of the curve. But move anything in the magenta end, and that's all you'll effect. You may move it closer to green, but until it crosses over into green territory, you will not add any green, and you will not effect the other pixels that are green.

    That one pixel MUST be magenta OR green. It cannot be both.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
    edited September 20, 2005
    You also cannot have blue and yellow on the same pixel. But wait, you say, that's green. Right. And that's handled by the A curve. You can have a bluish green or a yellowish green, but you do not mix blue and yellow to get green in LAB.
    Moderator Emeritus
    Dgrin FAQ | Me | Workshops
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited September 20, 2005
    This is an awful lot of theory for a discussion of Ch. 3. Ch. 2 is where it belongs really. Perhaps we can get a moderator to move this discussion?

    Anyway, in order to understand the relationship of LAB, RGB, and CMYK, we have to understand that the colors are actually defined in terms of one another.
    1. Red is a primary in light.
    2. Cyan is the pigment opponent of red. It is defined as the pigment that reflects green and blue perfectly but no red at all. So in light, cyan is composed of equal parts green and blue, but no red.
    3. Green is a primary in light.
    4. Magenta is the pigment opponent of green. It is defined as the pigment that reflects red and blue equally, but no green at all. So in light, magenta is composed of equal parts red and blue, but no green at all.
    5. Blue is a primary in light.
    6. Yellow is the pigment opponent of blue. It is defined as the pigment that reflects red and green equally, but no blue at all. So in light, yellow is composed of equal parts red and green, but no blue at all.
    Given these definitions, we can see why the pairs green, magenta and blue, yellow are called opponents. There can be no green at all where there is magenta, by definition. Shine a green light on a magenta surface and you see black; nothing is reflected. Magenta is defined in terms of what it doesn't have, namely green.

    To the extent that a color is created by combining magenta and green, that color just gets darker and that darkness can be expressed in the L channel instead of as separate values of magenta and green. You can experiment with this in the RGB color space. To get yellow, set the red and green values to 255 and red to 0. Now increase the blue. As you do so, the light does not actually become more blue, it becomes less yellow, paler and brighter. You cannot add enough blue to get a bluish yellow light. If you increase the blue all the way to 255, you get white light.

    That's the beauty of LAB. It is constructed from the fundemental definitons of the other two color spaces in a very pure way.

    And what does this have to do with anything anyway?
    If not now, when?
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 20, 2005
    DavidTO wrote:
    You also cannot have blue and yellow on the same pixel. But wait, you say, that's green. Right. And that's handled by the A curve. You can have a bluish green or a yellowish green, but you do not mix blue and yellow to get green in LAB.
    I think that our impasse is more semantic than anything else. One has to acknowledge that the a curve and the b curve merely represent values on a linear scale. So for a given pixel, you add or subtract value from that pixel in a linear way. Thus whether you number it 0 to 255 or make 0 neutral and have it go to -127 to +127, each successive value still has the same distance from its neighbors as does every other.

    So even with the stipulation that blue and yellow are modified completely independently, that doesn't change the fact that in the end you are still able to modulate only two parameters at any given time at the same pixel. If green and magenta are mutually exclusive, then you can't adjust them simultaneously at one pixel -- you can only modify one of them.

    In other words, you have not said that you can modify [green AND magenta AND blue AND yellow] at the same time: you are saying that you can modify [(green OR magenta) AND (blue OR yellow)]. So you have 4 choices in curves, of which you can modify only 2 for any given pixel:

    1. green/blue
    2. magenta/blue
    3. green/yellow
    4. magenta/yellow

    Again, if blue and yellow are mutually exclusive at the same point, and same with magenta and green, then you only get to choose 2 of them for a given pixel. By the same token, since you only get to move 2 points (an a point and a b point) for that given pixel, you're stuck with 2 color variables. Sure you can switch that pixel from green to magenta, but that doesn't add a new variable -- all it does is change how you're filling in that one variable.

    Now, in the end this probably doesn't matter at all, because you can use RGB to produce the same color as you produce in Lab (yes, I'm assuming a mutually shared portion of the color gamut). In fact I've found in my own experimentation that Lab has one huge advantage over RGB in color modification -- specifically, you don't alter tonal information when you adjust the color channels. You can get around this in RGB, though, if you do your curves modification on an adjustment layer that's set to 'color' blending mode. Another possible (though this should be subjective) advantage is that sometimes it's a bit more intuitive to use the color channels in Lab specifically because (as I, at least contend) there are fewer variables -- you can play with one color, but then neutralize a cast without switching channels. In RGB you have to go to the other channels to counter a color cast.

    So I apologize for making a long (though interesting) mathematical debate that doesn't amount to any practical difference in the end, though I think it really does help to understand what one is doing when one does it.
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • ginger_55ginger_55 Registered Users Posts: 8,416 Major grins
    edited September 20, 2005
    Not being as inclined to deep discussions as some seem to be, I just have a question:

    If my photo were one pixel, just one pixel: That is all, doesn't matter big or small, just one pixel,

    Would I be changing the pixel color as I did the lab curves, in any way I might do them?

    (I don't know why I am asking that, I don't know what else could be happening, but somewhere in my head that question is begging to be asked)

    ginger

    whose only real problem with this whole thing is the inability to understand what steepening means at any point. I can steepen a whole curve. I have no idea how to steepen a point on a curve.

    Rutt, you have tried to explain this to me, could PF, David, someone different give it a go. That would probably be most helpful to me in the long run. So basic.
    After all is said and done, it is the sweet tea.
  • snookman23snookman23 Registered Users Posts: 46 Big grins
    edited September 20, 2005
    Question and comments.
    rutt wrote:
    This is a really good question. Dan is rewriting Professional Photoshop and plans to publish in 2006. I think he has learned a lot from the LAB book. Basic color correction (the stuff of chapters 1, 3, and 4 in the LAB book) is really a LOT easier to do in LAB than in CMYK which is how Professional Photoshop starts out. I think that approach is probably obsolete and actually very difficult to learn. On the other hand, Professional Photoshop covers important stuff (sharpening) not in the LAB book.

    On balance, I'd get the LAB book and follow along until you lose it (odds are we all will at some point.) Wait for the next edition of PP (and the dgrin reading group that will go with it.)
    Thanks Rutt. You all have been very helpful so far. I am learning alot. The tutorials are great.
    I'll actually be scanning color negatives probably using a Nikon V Ed film scanner and VueScan. I'll have to work out the details of adjusting more using the scanning software or photoshop.

    As a side note, in the color theory discussion. I am not a graphic artist or color theorist, but an engineer. LAB seems to be a technical theory which, for someone with a technical mind, plots colors on a coordinate system. For the technical people, the A channel is the x-axis, the B channel the y-axis and the L channel the z-axis. Values can be negative or positive in the A channel (x-axis) and the B channel (y-axis), but only positive in the L channel (z-axis). For anyone losing me, think of a 3-d chess board. So, if a color is (20,20,20) we have all positive values. If it is (30,30,30) we still have all positive values. Is (20,20,20) more negative than (30,30,30)... NO. A number is ONLY negative or positive!! Same goes for colors. Basically, its semantic anyway, but you wouldn't describe 4 as less positve than 8.
    Remember folks, all of these color spaces are a human way to describe something that can be very vague. Everything in this world can be broken down to a mathematical description. Is it always right... No. That's why this is called color theory and not color law. Same with light, is it a wave or a particle.... well, it behaves like both, so both THEORIES are true.
    In my mind, LAB makes sense, its easy for me to visualize a three dimensional space and translate that to colors.
    Anyway, I ramble, thanks for everything.
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 20, 2005
    ginger_55 wrote:
    Not being as inclined to deep discussions as some seem to be, I just have a question:

    If my photo were one pixel, just one pixel: That is all, doesn't matter big or small, just one pixel,

    Would I be changing the pixel color as I did the lab curves, in any way I might do them?

    (I don't know why I am asking that, I don't know what else could be happening, but somewhere in my head that question is begging to be asked)

    ginger

    whose only real problem with this whole thing is the inability to understand what steepening means at any point. I can steepen a whole curve. I have no idea how to steepen a point on a curve.

    Rutt, you have tried to explain this to me, could PF, David, someone different give it a go. That would probably be most helpful to me in the long run. So basic.
    You can't steepen a single point, but you can change its individual value. The steepness at that single point is its calculus derivative (its instantaneous rate of change), but the derivative does not have anything to do with its individual value. So you can change the color of an individual pixel. Thinking in RGB terms, 0,0,0 is black and 255,255,255 is white. Just using 255 possible values per channel means that you can encode assign a single pixel more than 1.6 million ways (and that's just in 8-bit). Mathematically the same has to be true in Lab, simply because when you convert from Lab to RGB and back again you're not changing the color of any pixel (unless you clip off part of your color gamut).
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 20, 2005
    snookman23 wrote:
    In my mind, LAB makes sense, its easy for me to visualize a three dimensional space and translate that to colors.
    Anyway, I ramble, thanks for everything.
    In my own conception of it RGB chiefly differs from Lab in that the 3 channels all contain both color AND luminance information. In Lab luminance is managed as an independent channel, and color information is encoded by coordinates on 2 separate channels. I think the confusing aspect of RGB is the inseparability of color and tone. The confusing thing about Lab (for me) is the color logic, though it doesn't have to be completely understood to be put into practice.

    (Parenthetically, we live in an RGB world -- the cones in our foveas are sensitive either to red, green, or blue light, and both cones and moreso rods are sensitive to monochromatic brightness -- and these signals, R,G,B, and brightness are what get transmitted neurologically to our occipital cortex and constitute the data from which we perceive and interpret all visible colors).
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
Sign In or Register to comment.