Photoshop Lab Color: Ch. 1

rutt

This chapter describes the most cornerstone technique of LAB colorspace and then explains why it is effective and why it is eitehr much harder or impossible in other colorspaces.

This technique is:

1. Convert the image to LAB
2. Curves
3. Steepen A and B channels symetrically by bringing in the endpoints of each curve toward the center equally. After this, the curve (line actually) will still cross the center horizontal center at the vertical center.
4. Steepen the L curve through the areas where the detail is of most interest.
5. Apply the curves
6. Activate only the L channel
7. USM, trying the values 200, 1.0, 10

In the course of teaching this, Dan also points out perhaps the most important tool for writing good curves.

With the curves dialog open, you can mouse over the immage with the (left) mouse button held down. A point will appear on the curve showing exactly the point on the curve of that point on the image. By moving the mouse accross an area of interenst in the image, you can figure out which part of the curve controls that area.

Dan uses a few images of canyons and one of Yellowstone lake in the fog to illustrate. I'll repeat with an image of my own.



I took this just before 6pm. When I took it, it felt like early evening. But the image seems flat and there is little sign of that great sweet light that I remembered.

I followed Dan's recipe, steepening both the A and B curves by pulling in both endpoints by 16 points:





Now some of the color I remember is restored to the sky and the sand is warmer and, well, more sandy just as the dune grass is greener. But the image is still a lot paler than my memory of it. Using the mouse on the L curve, I discovered that the areas of interest are in the midtones and that my shadows aren't nearly at the top of the curve. Also my highlights aren't very close to the bottom of the curve. So there is lots of unused contrast which can be applied to the areas of interest in this image:





Just as the A and B curves increased the color differences, this curve increases the brightness differences, increasing the contrast. It's really the same principal

Lastly, I sharpened the L channel, using Dan's parameters:






Voila!

What's going on? For me the interesting part of this chapter is the theory behind it. Granted the technique is a super simple way of improving tons of shots, but understanding why it works opens the door to a greater understanding of how to use (and not misuse) LAB color corrections and of digital color correction in general. Human sight is a marviously complex and highly evolved system. It self calibrates to emphasize differences in color and in shade. When we are in the middle of the forest surrounded by infinite subtle shades of green, our vision pulls these shades apart allowing us to see enormous variation. On the beach in the late afternoon, we can discern traces of the impending sunset and see the complexity of hue in the sand and sky. But the camera doesn't do this. It doesn't "know" when to do it. When we look at a picture, our vision does not self calibrate because the picture only occupies a small portion of our field of view and is surrounded by all kinds of competing cues. So the picture won't capture the same intensity and variety of color we saw when we took it. Even if it is a faithful rendering of the "true" colors of the scene, it does not capture what we saw.

The basic LAB enhancement can restore the picture so that it captures the vision of our memory. Steepening the A and B curves moves colors apart from each other. This kind of symetrical curve steepening applies a constant multiplier to the colors. If it was X amount of green before, it will be X*N green after and similarly for magenta, blue, and yellow. The green gets greener in proportion to how green it was to begin with. The difference between blue and yellow is magnified as slight variations from neutral are pushed away from neutral toward in the direction of their tint. So the sky of my lighthouse shot becomes more of a tapestry of blue and red. It looks more like the sky just before sunset, more they way I remember it.

Steepening the L curve over the areas of interest in the image similarly apes the way our vision works. The steeper the curve, the greater the detail, as Dan likes to say. I found no interesting detail beyond darker than the point in the grass just to the right of the road. So I pulled the right end of the L curve in almost to that point. Then I steepened it through the midtones to bring up the detail in the sand, lighthouse, and sky.

Dan spends a certain amount of time comparing thte results of alternative non-LAB techniques for accomplishing the same thing. I don't want to recap these arguments in detail. If there is someone who really feels that it's important to be able to do this correction without using LAB, please raise the issue. Suffice to say:

1. This isn't going to be nearly as easy any other way.
2. Learning this technique will open the door for variations and different powerful uses of LAB. So why resist?

I guess there is one more thing worth mentioning. This 1st simple LAB technique works well for images which seem duller than our memories of the scene. The worse this difference, the better the technique can work and the more steepening is called for. But try applying to a picture that already has very bright distinct colors or to an image with were delicate subtle color variations are important (portraits) and the results won't be so fortuitous. Also if the image has a color cast, this technique will only make it worse.

OK, does someone else want to post some examples of using this technique? Have I missed or mistaken some points of theory?

Enjoy (LAB book discussion)

pathfinder

"I guess there is one more thing worth mentioning. This 1st simple LAB technique works well for images which seem duller than our memories of the scene. The worse this difference, the better the technique can work and the more steepening is called for. But try applying to a picture that already has very bright distinct colors or to an image with were delicate subtle color variations are important (portraits) and the results won't be so fortuitous. Also if the image has a color cast, this technique will only make it worse.

Very good discussion John, you have defintely set the bar high.

One thing not mentioned is that rather than steepening the a and b curves and spreading out the hue, flattening the a and b curves should flatten out the color also if colors were overdone.

rutt

Besides missing the part about curve flattening, I also missed a trick which Dan shows (off). Extreme A&B curve sharpening can bring out colors which aren't really there at all in the original. For example, here is a shot from this year's Boston marathon:



And here I have applied some some very extreme curves to get what Dan calls "the man from mars" effect:





Dan says that no client would want this and probably he is right since it's no longer the late 60s. But after blending this back into the original at 30%, I got this:



Now the blues introduced by the steep curves make interesting highlights on the too flat faces of the original, and perhaps some client would want.

Questions for discussion:

1. Could I have accomplished the same thing with less steep curves and no blend? Why?
2. This trick doesn't seem to work best with the darkest skinned people and not at all with light skinned people. Why?
3. The runner's shirts look awfully blue in the final version, probably too much for my client. Why? What could I have done about it?

rutt

Dan's basic philosophy
Here is one more topic for discussion. I think that people who don't buy into Dan's basic philosophy about photographic images versus vision are going to balk pretty quickly at his techniques. The discussion in Chapter 1 of Professional Photoshop covers this more directly than what I have read so far in the LAB book, but it's a pretty deep assumption in both books. In a nutshell:

The "truth" the camera captures is often only a pale shadow of our vision. Even the best shot images often fail to capture the colors and contrast of our vision. The objective truth the camera captures is very different from what we see. Good postprocessing can exegerate the colors and contrast (among other things) to bring the image closer to our original vision.

Opinions?

Khaos

rutt wrote:

The "truth" the camera captures is often only a pale shadow of our vision. Even the best shot images often fail to capture the colors and contrast of our vision. The objective truth the camera captures is very different from what we see. Good postprocessing can exegerate the colors and contrast (among other things) to bring the image closer to our original vision.

Opinions?

I agree completely. The brain deciphers what the eye gives it, so we don't always see what is truly there, thus we have those optical illusions.

Also, photos tend to, like music, have an emotional attachment to them. They talk to us through our eyes and if we need to adjust the colors and shadows to get the feeling across we should. Photography shouldn't be about getting a technically great shot every time, it should be about the photographers interpretaion of what he\she is shooting.

DavidTO

Thanks for the in-depth summary, Rutt. Looking forward to getting my copy and joining in...and I'm a bit nervous about meeting the high standards you've set in your summary.

rutt

DavidTO wrote:

Thanks for the in-depth summary, Rutt. Looking forward to getting my copy and joining in...and I'm a bit nervous about meeting the high standards you've set in your summary.

Don't worry. This chapter was the easiest chapter and the material was very familiar to me. Moreover, this wasn't even the first time I've written an explanation of this particular technique. Looking forward, I'm sure we'll all get out of our depth and the playing field will be level.

Right now, I'd like to see some practice with the basic technique from Ch. 1. Before and afters of shots were it really doesn't work are just as interesting as ones where it does.

DavidTO

How do you get your curves dialog to show percentage change instead of absolute value?

DavidTO

OK, here's my first.

Here's the curves I applied:



The straight RAW conversion:



And the result:

DavidTO

K, one more for now.

Original RAW conversion:



a and b curves:



The result:



Lightness curve:



And the final:



Now, I may be getting ahead of myself with this one. You'll notice that I didn't adjust the a and b curves symmetrically. This is one of those sweet light shots that we were talking about in another thread. I found the first to magenta, so when I adjusted the a/b curves, I just guessed about what I should be doing and got a color that I found more pleasing, but still seemed sweet to me. The original RAW conversion definitely would have left her looking sunburned. How'd I do?

I'm also not sure about the lightness adjustment I made. Is it too much? There's a very fine line between giving the image the pop it deserves and preserving the lovely softness of the light.

BTW: that's Lynnesite's horse Ember on a visit a year ago this month. Gotta back out to see her and her horses, and soon!

rutt

DavidTO wrote:

How do you get your curves dialog to show percentage change instead of absolute value?

It seems you figured this out. But for others, PS has this annoying property. If the curves are set up with darkness on the right, they show percentages, if the other way, they show absolute values. I wish these were independent options. Perhaps they are and I don't know how to control them.

ginger_55

how did you get percentages????


I did a whole bunch of photos, just because I needed to work some stuff up anyway. But I probably didn't do it right. Mine were not "off", they sometimes needed the punch I get in selective colors. If I really wanted a photo, I did that, too.

So I could not get the percentages.
I just subtracted whatever the number was on the page.

And I would need a whole tutorial on how to show those charts and things, but I can show you some photos worked up.

I will show you one from RAW, just LAB and with LAB and selective colors.

ginger

ginger_55

These are all full frame, no cropping was done
from RAW




From LAB




From RAW plus selective colors with about +10 in both red and yellow




I didn't have any where it didn't work, what I had were a few where I wanted a little punch in certain areas, such as above and below.

RAW below




Just LAB below:




LAB below plus selective colors w about +10 in red. Then I used the burn tool on her lips and eyes, along with the sharpening tool, lightly on her eyes and hairl

ginger_55

I wanted to see how this would work. I am trying to better my bird shots since I seem to like them all by myself. This is a shot of a bird in front of a decrepit shack that I love. I worked it up last week, it was not the best, but I thought maybe it could be better.

This was last week, CMYK black channel, but no LAB:




This was today with just LAB





This was the same as above with, again, selective color added as punch in the yellows and red. I use yellow a lot with grass and with a bird w yellow on it. I use red to enhance wood and grass.



Below is a bird in flight, I wanted to see what would happen there. This is with just the RAW below:




This was the above bird today w LAB plus selective colors upped the yellow, plus I added a cooling filter which improved the sky, I thought.




All full frame again.

ginger

rutt

Ginger, it's hard to understand what you did. Please show your work as David and I did.

rutt

David, the bird picture works, but I might have been braver with the A+B steepening.

With the horse picture, you definitely did skip ahead a little too fast. You have introduced a color cast. The neutral horse is now yellow and to me looks like he belongs in the merry-go-round at Disney Land. That's the reason for keeping the center point of the curve in the center: it prevents neutral colors from acquiring a cast. In this case, I don't think there was a cast, but the child's face is too light and may need a little less magenta and/or more yellow. Try steepening the L curve through the area of the girl's (and horse's) face, the highlights. Try splitting the difference and making the A curve a but flatter and the B curve steeper untl you get good CMYK readings on her flesh (Y >= M).

Soon, you'll find out how to do better, but for now keep the curves symetrical.

DavidTO

rutt wrote:

With the horse picture, you definitely did skip ahead a little to fast. You have introduced a color cast. The neutral horse is now yellow and to me looks like he belongs in the merry-go-round at Disney Land. That's the reason for keeping the center point of the curve in the center: it prevents neutral colors from acquiring a cast. In this case, I don't think there was a cast, but the child's face is too light and may need a little less magenta and/or more yellow. Try steepening the L curve through the area of the girl's (and horse's) face, the highlights. Try splitting the difference and making the A curve a but flatter and the B curve steeper untl you get good CMYK readings on her flesh (Y >= M).

Soon, you'll find out how to do better, but for now keep the curves symetrical.

The flesh tones definitely had too much magenta in relation to the yellow, that's why I did that. This shot was taken just after the sun had set behind a hill, so it was very warm.

When I steepened the L curve in her face, it got too contrasty and harsh.

I guess the part that's getting ahead is exactly how to give her more yellow/less magenta in LAB.

ginger_55

Rutt, I have no idea how to show what I did. I never have known how to show my work space, and I have commented about that in the past and asked for help.

However, I know what I did, I know what you all did. It is working for me. Once I get started, I just keep working things up, unless there is a problem.

If someone wants to show me how to show a work flow, I will, but not often, and not right now.

I would probably do it once. It is quite simple though. I printed out the pages of your tutorial and followed them, as best as I could, with each photo. The only thing I did differently was to use the numbers and subtract, as I didn't know how to get the percentages.

ginger

PS, it is OK, you can just ignore my stuff up there. I am pleased, smile.

gus

I have always ignored curves
Until David sent me 'the photoshop CS2 book for digital photographers' by scott kelby.

I couldnt believe how flat this shot was i took yesterday so i sat down with the book & took it through the 18 steps on colour correcting using the darkest & lightest points (the 2 sheds in the foreground)....only took about 5 mins.

No massive difference but its something that improved the shots colour. Got rid of that purple look also.

Thanks David

Before




After

DavidTO

Gus,

Check out this thead and try it out on that shot.

gus

Hence my issues mate...whats LAB ? I have to have the book open & follow it through step at a time.

DavidTO

Humungus wrote:

Hence my issues mate...whats LAB ? I have to have the book open & follow it through step at a time.

LAB is a color mode, like RGB.

RGB: Red, Green, Blue
CMYK: Cyan, Magenta, Yellow, Black
LAB: Lightness, A, B....yeah, I know the logic falls apart there.

But read the first post in the thread, it should tell you everything you need to know. If not, ask a question.

You enter LAB mode by menu Image>Mode>LAB. When you're done, go back to RGB Image>Mode>RGB.

DavidTO

Here's a rough try with your M image.

I pulled in the A+B curves 15 points on either end (85/15) and adjusted the curves on the Lightness channel slightly.

gus

DavidTO wrote:

Here's a rough try with your M image.

I pulled in the A+B curves 15 points on either end (85/15) and adjusted the curves on the Lightness channel slightly.

Ok...you convinced me...i will give it a burl & report back.

DavidTO

Rutt,

Another question I have is, how do you ensure that you aren't making colors that aren't real? I know that LAB can make colors that don't really exist. I've tried proofing with sRGB and CMYK, but I don't really understand how the Gamut check thing works. When I change the mode back to RGB do I automagically lose the colors that RGB can't represent? That would make sense, since that's my understanding of what happens if you go through CMYK; it's lossy.

It just seems that I should be able to make colors in LAB that are so vibrant as to be beyond the gamut of RGB and, of course, CMYk.

gus

OK got the curves worked out (L being for lightness i assume) but where is the USM ?

DavidTO

Humungus wrote:

OK got the curves worked out (L being for lightness i assume) but where is the USM ?

That's Unsharp Mask, under sharpen in the Filter Menu. I actually use Smart Sharpen in CS2, but either is fine. And you should probably stick with Rutt's recommendations for now, he understands this much better than I do.

gus

Hmmm...any better ?

DavidTO

Humungus wrote:

Hmmm...any better ?

To my eye you could be more aggressive with the A+B curves, making the color more intense.

DavidTO

ginger_55 wrote:

Rutt, I have no idea how to show what I did. I never have known how to show my work space, and I have commented about that in the past and asked for help.

Ginger, you may want to start another thread asking the Windows folks (can't help you, I'm a Mac guy) how to do screen and window captures. It should be fairly easy to do.

pathfinder

DavidTO wrote:

The flesh tones definitely had too much magenta in relation to the yellow, that's why I did that. This shot was taken just after the sun had set behind a hill, so it was very warm.

When I steepened the L curve in her face, it got too contrasty and harsh.

I guess the part that's getting ahead is exactly how to give her more yellow/less magenta in LAB.

Dave, the background shifted from a neutral grey to a greenish cast also behind the horse. The flesh tones seemed ok, but their def was a color shift.

I tried John's recipe taken from chapter 1 on a wedding shot I did last weekend, just for grins to see what effect it would have on people rather than vegetation or canyon walls like Margulis talked about in chapter 1.

John's post:
"This technique is:

1. Convert the image to LAB
2. Curves
3. Steepen A and B channels symetrically by bringing in the endpoints of each curve toward the center equally. After this, the curve (line actually) will still cross the center horizontal center at the vertical center.
4. Steepen the L curve through the areas where the detail is of most interest.
5. Apply the curves
6. Activate only the L channel
7. USM, trying the values 200, 1.0, 10 ""


This is exactly the recipe I used for the second image I am posting here. I did use a slightly different amount of USM as I was shooting 16 Mb files - so I used 200, 2.0 10 for sharpening




And the shot with the increase in the a and b curves - straight lines through the center, but one block steeper along the top and bottom row, minimal changes in the L channel.



Definitely more color in the flowers, the vegetation, and the skin tones. I was surprised by how much more I liked the second shot. The first seems to have a faint grey cast to it by comparison.