Photoshop LAB Color: Chap. 3
DavidTO
Registered Users, Retired Mod Posts: 19,160 Major grins
Chapter 3 is about preparing for expanding on the benefits of LAB that were introduced in the symmetrical adjustment recipe.
REVIEW:
The chapter starts with a review of that recipe. LAB separates color and contrast. After steepening the A and B curves to improve the range of colors in the image, adjusting the L curve involves finding the range that most needs the adjustment. While in L curves, mouse over the image (click and drag) and notice the area of the L curve that is activated. This is the area that you need to steepen. You can do this with a simple curve that steepens that area. Sharpening is then applied to the L channel.
Images that have L curve problems (too dark, too light, etc.) also tend to be colorless. You can't just lighten a dark image, because chances are it's too monotone, as well. You need to increase the range of colors as well.
NEW CONCEPT:
The next element that Margulis adds is this: steepening the A curve will intensify the green and magenta, the B curve the blue and yellow. You can use this to control the color balance. As long as you keep the curve straight at this point, and passing through the center point, the image will remain otherwise neutral.
One example he gives is for images that are predominantly green foliage. If you increase both A and B equally, you'll end up with greens that are a bit too yellow. Simply reduce the steepening on the B curve, and you'll reduce the yellows. Keep in mind that this will also decrease the blues.
In the Chapter 1 thread, I got ahead of myself and was trying to correct this image of my daughter. The skin tones are too magenta to start with. I could tell this by the numbers by checking the CMYK values, knowing that unless you want the skin to look sunburned yellow should always be greater than magenta.
When I increased both A and B equally, the magenta in the skin tones got even stronger. Here is the result of steepening both equally:
If I apply different curves to A than to B (and in this case I actually made the A curve LESS steep), I reduce the magenta, increase the yellow and produce more pleasing skin tones:
Here are the curves I applied:
You can see that reducing the steepness of A and increasing it in B made the image overall more yellow than magenta, but it kept the blue and yellow balanced.
For the purpose of going over this new tool that we have, I'm ignoring the L curve (which I modified as well) since in LAB the colors are independent of Luminosity.
This is a powerful, new overall color correction that we have.
Margulis spends a bit of time talking about our perception of the greens of nature and skin tone. These are two situations where our mind sees most differently from the camera. We're sure that the greens are richer and more detailed than they really are. And we are also sure that skin is more tanned, bronzed than it really is. His argument for these two situations is that we should satisfy our perceptions of these types of images, not the reality. If we are too literal with greens or skin, our perception of the image will that it is flat and washed out.
IMPORTANT TERMS:
Full Range. This means that the lightest and darkest parts of the image are handled right. Appropriately is the word he uses. This is also called setting highlight and shadow, which we're all familiar with. In RGB and CMYK achieving a neutral color is a matter of balancing the numbers. Neutral colors in RGB have the same value for each of the channels. In CMYK, the channels that need to be the same are magenta and yellow, with a bit more cyan.
In LAB you can set your highlight with the L channel and then just make sure that you have a neutral color by making sure that A and B both are 0. This means that you can make huge changes very quickly in LAB. This is because L is, as Margulis calls it, a bull. It's a very powerful tool. It can correct serious problems with contrast. This also means that RGB and CMYK will handle some situations better, since they are not a bull in a china shop. Those instances will be covered later in the book.
The point of "Full Range" is to "maximize range while retaining highlight and shadow detail." The point here is very clear: be cautious, especially if you will be doing final work in RGB or CMYK, where you can further adjust the highlights and shadows. Once you've lost highlight or shadow information it's gone, never to be regained. So treat it with care.
Impossible colors are colors that just aren't right. You don't always have the luxury of a neutral color to balance from, but understanding LAB numbers (as we'll all get better at as we work through this), you can think through an image and make intelligent choices. The large the A or B number (positive or negative) the more color. The closer those values are to 0, the more neutral they are.
What I talked about above in setting the L channel curve; that's called Allocation of Contrast. This concept basically says that you should give contrast (tonality range) to the part of the curve where the image lives. Or, more specifically, the part of the image that you want to focus the attention of the viewer on.
Margulis goes through two examples at this point. A dark, dense forest and head shot of a model. His basic point here is that the way we see these scenes is different than the camera. We see more in the shadows, more shades of green, more intense green, and a tanner face. Our job as post-processors is to not capture the scene accurately from a scientific standpoint, but from the way that we think we see. We have localized contrast and a much wider tonal range available to us as we walk through this world. We need to overcome the limitations of the photographic image by making the image optimized to reproduce our memory of the scene.
MY MNEMONIC DEVICE:
Understanding the L curve is easy. Dark is dark and light is light. The A and B curves are more difficult to keep straight. Once you get in your head that green is opposite of magenta and that blue is the opposite of yellow you'll know which colors are paired. After that just remember B = Blue. Then you'll know that the B curve is blue and yellow. You could also add: A = mAgentA. There's no "a" in blue or yellow.
Another thing to get clear on is that in LAB positive values are warm (magenta and yellow) and negative values are cool (green and blue). I know that yellow and green can't be opposites; they are too closely related. That helps me remember that yellow is paired with blue.
Anyway, it helps me.
NUTS AND BOLTS:
It's time to make sure that we're all working on the same page. Some of the settings that Margulis recommends are important, others are just what he uses. His suggestion is to do as he does, since it will lead to less confusion when reading the book.
First, the eyedropper is set by default to point sample.
This is not very useful since it could pick up a bit of noise or dust. When you've got the eyedropper tool selected you can change this in the upper left. 3 by 3 or 5 by 5 is fine.
The info palette is also very important. There are little eyedroppers next to the colorspaces on each side. This is a drop down menu that you can change. The left side is set by default to Actual Color. Leave it as is. The right side you should set to the colorspace that you are most familiar with and understand the numbers best. Margulis says that for him that is now LAB, but for most of us (especially those of us who have to read this book!) we'll find RGB or CMYK most natural.
Having said that, there are different RGB and CMYK colorspaces, all of which will yield different results. For the purpose of working with this book he recommends that you set your RGB to sRGB and your CMYK to U.S. Web Coated (SWOP)v.2. You can change these in your Color Settings, found in the Edit menu.
In order to change between RGB and LAB, Margulis uses the Convert to Profile command. His reasons are pretty esoteric and I'm pretty sure that we can all just use the Image>Mode menu.
If you want to use the Curves dialog eyedroppers to set values automatically (black and white points), he suggests the following settings:
Black: 6L 0A 0B
White: 97L 0A 0B
You set these numbers by double clicking on the eyedroppers in the Curves dialog box and changing the values for LAB there.
Extra Credit:
We all see color differently. Especially color-blind people, who cannot distinguish the green/magenta very well. This section of the chapter is really only useful if you read it and study the images. The interplay of A and B is so crucially important to what we do, that as much experience you can get in playing with the numbers in your head as you look at images is important. Take a look at the images that he's put in this part of the chapter, read along and you'll gain more insight into how A and B work together.
REVIEW:
The chapter starts with a review of that recipe. LAB separates color and contrast. After steepening the A and B curves to improve the range of colors in the image, adjusting the L curve involves finding the range that most needs the adjustment. While in L curves, mouse over the image (click and drag) and notice the area of the L curve that is activated. This is the area that you need to steepen. You can do this with a simple curve that steepens that area. Sharpening is then applied to the L channel.
Images that have L curve problems (too dark, too light, etc.) also tend to be colorless. You can't just lighten a dark image, because chances are it's too monotone, as well. You need to increase the range of colors as well.
NEW CONCEPT:
The next element that Margulis adds is this: steepening the A curve will intensify the green and magenta, the B curve the blue and yellow. You can use this to control the color balance. As long as you keep the curve straight at this point, and passing through the center point, the image will remain otherwise neutral.
One example he gives is for images that are predominantly green foliage. If you increase both A and B equally, you'll end up with greens that are a bit too yellow. Simply reduce the steepening on the B curve, and you'll reduce the yellows. Keep in mind that this will also decrease the blues.
In the Chapter 1 thread, I got ahead of myself and was trying to correct this image of my daughter. The skin tones are too magenta to start with. I could tell this by the numbers by checking the CMYK values, knowing that unless you want the skin to look sunburned yellow should always be greater than magenta.
When I increased both A and B equally, the magenta in the skin tones got even stronger. Here is the result of steepening both equally:
If I apply different curves to A than to B (and in this case I actually made the A curve LESS steep), I reduce the magenta, increase the yellow and produce more pleasing skin tones:
Here are the curves I applied:
You can see that reducing the steepness of A and increasing it in B made the image overall more yellow than magenta, but it kept the blue and yellow balanced.
For the purpose of going over this new tool that we have, I'm ignoring the L curve (which I modified as well) since in LAB the colors are independent of Luminosity.
This is a powerful, new overall color correction that we have.
Margulis spends a bit of time talking about our perception of the greens of nature and skin tone. These are two situations where our mind sees most differently from the camera. We're sure that the greens are richer and more detailed than they really are. And we are also sure that skin is more tanned, bronzed than it really is. His argument for these two situations is that we should satisfy our perceptions of these types of images, not the reality. If we are too literal with greens or skin, our perception of the image will that it is flat and washed out.
IMPORTANT TERMS:
Full Range. This means that the lightest and darkest parts of the image are handled right. Appropriately is the word he uses. This is also called setting highlight and shadow, which we're all familiar with. In RGB and CMYK achieving a neutral color is a matter of balancing the numbers. Neutral colors in RGB have the same value for each of the channels. In CMYK, the channels that need to be the same are magenta and yellow, with a bit more cyan.
In LAB you can set your highlight with the L channel and then just make sure that you have a neutral color by making sure that A and B both are 0. This means that you can make huge changes very quickly in LAB. This is because L is, as Margulis calls it, a bull. It's a very powerful tool. It can correct serious problems with contrast. This also means that RGB and CMYK will handle some situations better, since they are not a bull in a china shop. Those instances will be covered later in the book.
The point of "Full Range" is to "maximize range while retaining highlight and shadow detail." The point here is very clear: be cautious, especially if you will be doing final work in RGB or CMYK, where you can further adjust the highlights and shadows. Once you've lost highlight or shadow information it's gone, never to be regained. So treat it with care.
Impossible colors are colors that just aren't right. You don't always have the luxury of a neutral color to balance from, but understanding LAB numbers (as we'll all get better at as we work through this), you can think through an image and make intelligent choices. The large the A or B number (positive or negative) the more color. The closer those values are to 0, the more neutral they are.
What I talked about above in setting the L channel curve; that's called Allocation of Contrast. This concept basically says that you should give contrast (tonality range) to the part of the curve where the image lives. Or, more specifically, the part of the image that you want to focus the attention of the viewer on.
Margulis goes through two examples at this point. A dark, dense forest and head shot of a model. His basic point here is that the way we see these scenes is different than the camera. We see more in the shadows, more shades of green, more intense green, and a tanner face. Our job as post-processors is to not capture the scene accurately from a scientific standpoint, but from the way that we think we see. We have localized contrast and a much wider tonal range available to us as we walk through this world. We need to overcome the limitations of the photographic image by making the image optimized to reproduce our memory of the scene.
MY MNEMONIC DEVICE:
Understanding the L curve is easy. Dark is dark and light is light. The A and B curves are more difficult to keep straight. Once you get in your head that green is opposite of magenta and that blue is the opposite of yellow you'll know which colors are paired. After that just remember B = Blue. Then you'll know that the B curve is blue and yellow. You could also add: A = mAgentA. There's no "a" in blue or yellow.
Another thing to get clear on is that in LAB positive values are warm (magenta and yellow) and negative values are cool (green and blue). I know that yellow and green can't be opposites; they are too closely related. That helps me remember that yellow is paired with blue.
Anyway, it helps me.
NUTS AND BOLTS:
It's time to make sure that we're all working on the same page. Some of the settings that Margulis recommends are important, others are just what he uses. His suggestion is to do as he does, since it will lead to less confusion when reading the book.
First, the eyedropper is set by default to point sample.
This is not very useful since it could pick up a bit of noise or dust. When you've got the eyedropper tool selected you can change this in the upper left. 3 by 3 or 5 by 5 is fine.
The info palette is also very important. There are little eyedroppers next to the colorspaces on each side. This is a drop down menu that you can change. The left side is set by default to Actual Color. Leave it as is. The right side you should set to the colorspace that you are most familiar with and understand the numbers best. Margulis says that for him that is now LAB, but for most of us (especially those of us who have to read this book!) we'll find RGB or CMYK most natural.
Having said that, there are different RGB and CMYK colorspaces, all of which will yield different results. For the purpose of working with this book he recommends that you set your RGB to sRGB and your CMYK to U.S. Web Coated (SWOP)v.2. You can change these in your Color Settings, found in the Edit menu.
In order to change between RGB and LAB, Margulis uses the Convert to Profile command. His reasons are pretty esoteric and I'm pretty sure that we can all just use the Image>Mode menu.
If you want to use the Curves dialog eyedroppers to set values automatically (black and white points), he suggests the following settings:
Black: 6L 0A 0B
White: 97L 0A 0B
You set these numbers by double clicking on the eyedroppers in the Curves dialog box and changing the values for LAB there.
Extra Credit:
We all see color differently. Especially color-blind people, who cannot distinguish the green/magenta very well. This section of the chapter is really only useful if you read it and study the images. The interplay of A and B is so crucially important to what we do, that as much experience you can get in playing with the numbers in your head as you look at images is important. Take a look at the images that he's put in this part of the chapter, read along and you'll gain more insight into how A and B work together.
0
Comments
Thank you, and thanks for the help you gave me with it.
Dgrin FAQ | Me | Workshops
Here's the original RAW conversion:
If I apply equal changes to the A and B curves
and this L curve:
I get this:
So, following Margulis' logic I lowered the steepness of the B curve to compensate for the yellow in the green:
to get this:
My question is this: to my eye the even changes in A and B, the more yellow version, looks better, as this is evening sweet light. In addition, there's not much skin tone, but my son's hand there gets too magenta in the second version.
Are there times that it would be desirable to leave the accentuated yellow in the greens of nature? Which do you all prefer?
Dgrin FAQ | Me | Workshops
But you are right about one huge thing. Every image is different. What Dan really teaches is not recipes, but how to think about photographic color. I think you are on your way.
Am I getting addicted to LAB, going too far with it?
One thing I've noticed is this: I played with an image and then saved my curves (a very handy thing to do). I then did some other things and wanted to go back and grab the curves to post here. So I reopened the image (already corrected) and reapplied the same curves (effectively doubling them) with no intention of saving the image like that, it was just temporary to grab the curves. The image definitely looked overly saturated and processed. I went about my business and then canceled out of the curves dialog box. The image went back to what I had previously thought was darn good, and it looked washed out and boring. Just for a second...my eyes adjusted and I realized it was still darn good. But in contrast to the overly strong colors I had just been looking at, my eyes were fooled. Have to be careful of that...
Anyway, I can't bring myself to bring the curves down as much as you are suggesting...
Here's the image again with A and B the same, at 15. What do you think of this?
My previous favorite, for reference:
Dgrin FAQ | Me | Workshops
Anyway, I wanted to do my own experiment with LAB vs. RGB.
The first image is RAW conversion, the second is LAB and the third is my attempt to match the LAB in RGB. Curves below that.
LAB curves:
RGB curves/saturation:
My take? LAB looks better. Maybe that's partly because of my lack of experience overall, but that's part of the beauty of LAB. Powerfully good changes, relatively easily. To my eye, the orange rock in the RGB starts to look pretty day-glo. And the detail that I could easily pull out of the rocks with L curves are much better, IMO.
If anyone wants to try this themselves, the original file is here.
Dgrin FAQ | Me | Workshops
Here is my first go, on an image (glacier near Fafleralp, Switzerland) with just best effort RAW conversion (RawShooter Essentials).
Here is the image with the clouds and glacier enhanced with the L curve and colors boosted by a,b 20-80 steepening.
Here is the image the same L curve but colors boosted by a 20-80 and b 5-95 steepening, to keep yellow out of the greens.
Thanks!
Dan
"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
Very nice. Did you save your curves so we can see what you did? I'm mostly interested in the L, since you basically told us about the A+B. What I would like to see is if you could eek out any more of that shaft by tweaking the steepness of the L curve right there. I didn't even notice in the first. I saw the patch of light, but glanced right over the shaft. It is most excellent, and I'd like to see it a bit more pronounced.
What do you think?
Dgrin FAQ | Me | Workshops
The shaft of light is at about L = 25--30%, in the same region as many of the clouds. Because of the yellow in the shaft of light, maybe something more targeted could enhance it? It is more prominent in the printed version. Thanks for any ideas.
Dan
"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
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.
So I had this image before these LAB curves:
Here are my LAB curves:
Notice I got a tiny bit ahead of myself. My A and B curves aren't quite symetrical in order to get the snow and sky to be slightly blue or neutral instead of slightly yellow or neutral. This is Ch. 4 stuff.
Question,
Im just getting into the color correction business. I have been reading along with these LAB color posts and am following along.
I have yet to buy the lab book or Margulis' Photoshop book.
What would you recommend:
1. Study the Photoshop color adjustment book first and learn RGB adjustments and then study LAB
or
2. Dive right into lab.
Im not having any trouble understanding the numbers part or the concepts behind them.
Thanks
Dgrin FAQ | Me | Workshops
That Dan's books are all but trivial. They are NOT "Color Correction 101". If you dive into them headfirst without actually understanding the basic principles of PS/PSP you'll be up the creek big time. Of all the PS-related books that I own Dan's are the most advanced. I still have hard time sometimes going trough them (didn't finish them yet), but I think it would be many times harder or next to impossible if I was not pretty comfortable with PS in the first place by the time I got to them.
Don't let this passage above to disappoint you, however. It's meaning is quite the opposite: it's a fair warning. And you know what they say: "Forewarned is forearmed". :-)
HTH
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.)
In case you were happily nerding on on the vitrues of RGB vs LAB, I had Andy move this all to the Chapter 2 discussion here: http://www.dgrin.com/showthread.php?t=9739
Let's review and critique together.
I spent last weekend pretty much a bachelor, and spent most of it just experimenting with LAB on images. I learned a ton. I posted some of it. Like anything else learning LAB takes practice, practice, practice.
Like my kids' math teachers say, "Show us your work!"
Dgrin FAQ | Me | Workshops
Mark Twain
Some times I get lucky and when that happens I show the results here: http://www.xo-studios.com
Also, fwiw, I adjusted smugmug's autoadjust software to create more saturated colors and customer satisfaction definitely went up except in the case of portraits where skin had imperfections. Most people really like vibrant colors.
ginger
Thanks Baldy. I agree. The second one in my post (first I did) was too much. The first in my post (second version I did) is better. Sure is fun playing with LAB and figuring out what is too much!
Dgrin FAQ | Me | Workshops
Ginger, the second is more saturated and the greens are more bluish than the yellowish greens in the first.
Dgrin FAQ | Me | Workshops
As I said about this originally, all the post-LAB versions are 100x improved from the original. The differences between the various steepened versions is relatively subtle. I guess that this particular example didn't require less B than A steepening because in the original the grass was a touch on the green (vs yellow) side and so could be equally steepened, ending up greener than yellow. Or perhaps the "sweet light" was sort of magenta (vs yellow), meaning that the green in the grass needed more help than the yellow in the grass. David, take a color sampler reading on the original and find out. What are some LAB readings on the grass?
LAB curves are the biggest driver in your golf bag. You can get the ball almost to the hole most of the time with just a few seconds of work with a little practice and understanding. The point of this chapter is really to alert you to two things which can go if you insist on steepening both A and B curves the same: vegetation can look too yellow and fleshtones can look too magenta. When you see a lot of vegatation or a portrait, keep this in mind and use the color sampler tool to check whether you've pushed some important color to the point of being implausable. And now you have some ideas about what to do about it.
Dan is trying to teach us to think about color. The recipes are not hard and fast rules, but rather a structure. The book started out teaching a very simple and highly constrained recipe but as we get through chapters 4 and 5 the restrictions will mostly fade away and be replaced with understanding. Without that understanding, we have to adhere ridgely to simple rules; once we gain it, we can effeciently attack color issues directly.
This is true. Yet the difference between the two IS noticeable, and if Ginger's not seeing it, then something is wrong.
I will when I get a chance. It's my daughter's birthday today (yay!), so it may not be till tomorrow.
Dgrin FAQ | Me | Workshops
This was shot at a claaical music concert at Jordan Hall in Boston. It came out very warm, even setting ACR to "tungsten" didn't get close. I used ACR's WB eye dropper to balance on a point on the turtleneck. The result was this:
Looks, OK, eh? But I measured slightly higher A than B almost everywhere on her face. Typical values were A=16, B=14 and A=12, B=11. I couldn't find the reverse anywhere.
So I steepened the B curve symetrically enough to reverse this balance.
While I was at it I moved the light endpoint of the L curve inward just a touch to expand the contrast range.
That kind of skin tone problem is something that I've really been concentrating on lately--because I used to be fooled by it everytime. Before I looked at the numbers (first CMYK and now LAB), to my eye it looked natural. And the monitor tends to be much more forgiving with this than the print.
I'm amazed now how many times seemingly natural skin tone is improved by looking at the numbers, doing a little thinking and adjusting accordingly.
Nice work on a difficult shot.
Dgrin FAQ | Me | Workshops
I took 9 readings in various parts of the grass in my original JPEG that was converted from RAW.
The average was L39 A(16) B22.
Dgrin FAQ | Me | Workshops
This is an awesome discussion. By total coincidence, I picked up the LAB book yesterday and found this discussion via Google last night.
As I was playing with my own images, I discovered something really, really helpful to me.
When I'm making the curve adjustments, I found it really useful to make a separate curve adjustment layer for each type of curve adjustment (one for L, one for A, one for . This lets me independently see the effect of each adjustment without getting their influence all intertwined. At any point in time, I can also toggle one of the adjustment layers on/off to see how much of an effect it has. And, when I'm all done if I feel like I've slightly overdone one of the individual adjustments, I can easily just modify the opacity on that one adjustment.
Of course when you go back to RGB, you have to collapse all the adjustment layers since their adjustment can't be similarly expressed in RGB (the whole reason you're in LAB in the first place), but it's incredibly useful while you are making your adjustments.
I also find it valuable to make a seprate curve adjustment layer for each part of the image that I'm trying to make an L adjustment on. I find this much easier than trying to put all the adjustments into one curve. Furthermore, if you find that fixing the L curve for one part of the image messed up another part, you can easily just mask off the areas you don't want to affect if you have isolated each L adjustment area into it's own curve layer.
Here's the layers palette from one of my images:
And here's the image before (straight out of the camera):
And after fixing it with LAB adjustments:
I'm still deciding whether I've overdone it or made it a bit too blue, but wow this is powerful.
--John
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Great ideas, John.
I think it looks great, save the fact that I would remove the blue cast on the rocks to the right.
EDIT: on looking at it again, I think that the rocks on the right and the trail are a bit too magenta? The trail has a bit of blue, too, doesn't it? I haven't measured, just eyeballed. I would bring the magenta back, and maybe go a touch more yellow. I believe that would feel more realistic. Just my opinion .
Dgrin FAQ | Me | Workshops