Chapter 4, Professional Photoshop, 5th Edition

[gefillmore]

Chapter 4

Color and Contrast by LAB Mode: the ‘Ruttimentary’ Steps

(for much more advanced text, examples and discussion, please see Dan Margulis’ book, Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace, and/or the discussion thread (http://digitalgrin.com/showthread.php?t=18203) at Digital Grin)

The LAB mode is best used when you have a photo with drab, dull colors. Using LAB will drive the colors apart. Also LAB mode will treat colors and contrast separately, which will be seen to be very handy as we move into the chapter.

Photographs that have bright and different colors are not the best subject for improvement by the LAB mode. As stated, LAB mode drives colors apart and these types of images are already there.



http://gefillmore.smugmug.com/photos/119305947-M.jpg
Not for Lab mode adjustment

I have chosen a photograph of a barn I took just for the purpose of demonstrating the benefits of the LAB mode:

http://gefillmore.smugmug.com/photos/121332555-M.jpg
Jpeg copy of the raw file

First, I have attempted to improve the picture with the standard levels and curves in RGB mode (I usually try to do this intuitively. For a more formal method, see DavidTO’s tutorials- pop tute (http://dgrin.smugmug.com/gallery/2292454/1) – on ‘popping’ the photo in the How To section of the Digital Grin Forum).

I have positioned this image just above the LAB mode improved images for comparison purposes.

Now I am going to adjust the color of the photo in LAB mode. Go to Image: Mode>Lab color. Then open Image: Adjustments>Curves. Your first screen is the Lightness channel. Photoshop default for LAB mode curves is to have black input, output at 0, 0. Left click on the white triangle in the black-to-white bar underneath the graph and that will switch it to white input, output at 0, 0. This will put you in synch with Margulis’ writings and examples (and ours on Dgrin).

Now, click on the dropdown where it says Lightness and click on the a curve. We will go back to the Lightness curve after the a and b curves. Simply put, you move the top and bottom endpoints in by equal amounts, making sure the curve stays over the center point. The center point is a neutral area (white, black, or neutral gray). If you move the curve at this point, you are getting into adjusting for color casts. Not that this is an invalid adjustment, we’re just keeping it simple right now and making a basic adjustment on an image.

Please see the curve image below for a and b adjustments. This image has not been adjusted in LAB mode for contrast yet.

After levels and curves adjustments in RGB mode
http://gefillmore.smugmug.com/photos/119306425-M.jpg

http://gefillmore.smugmug.com/photos/121331693-M.jpg
After LAB mode curves (a and b) adjustments

Now for the Lightness curve, which deals with contrast.

You more or less curve this to taste. You can use the cursor on the photo to tell where on the curve that particular area of the photo will be affected when you move the curve. This is very sensitive and you have to adjust incrementally. To improve the photo, you usually want to steepen the curve at the points corresponding to the portion of the photo where you desire more contrast.

http://gefillmore.smugmug.com/photos/121331756-M.jpg
After Lightness curve adjustments- note the steepening of the curve in the lower half deals with the sky (lighter portion of photo) while the steepening in the upper half affects the barn and field (darker portions)-

http://gefillmore.smugmug.com/photos/119306740-M.jpg
LAB mode curve adjustments for barn image

You can sharpen the image by going to the Channels palette, clicking on the Lightness channel, then opening up Filter: Sharpen>Unsharp mask. For Amount, use 200%. For Radius, use 1.0 pixels. And for Threshold, use 10 levels. What type of sharpening you use depends much on the image, the resolution, what your tastes are, your audience.

http://gefillmore.smugmug.com/photos/121333604-S.jpg


http://gefillmore.smugmug.com/photos/121333409-M.jpg
LAB mode sharpened-unsharp mask

Then click on Lab in the channels palette to bring the image back to color.


Lab mode adjustment and sharpened

You almost always want to put the image back into RGB mode before saving, printing, or moving to the web (Image: Mode>RGB color).

Remember, LAB adjustments will increase the differences between colors, but it is only worthwhile if the differences are not too big to begin with.


Color casts-

As stated previously, the center point is neutral (white, gray, or black). However, if you have a particular cast to your photograph you can correct it by moving off the center point. In the image below, the 'white' is not. By a minor adjustment to the LAB curves, the red/yellow cast to the moisturizer bottle and sheet can be removed without greatly affecting the other colors.

In the Lightness curve there is a basic adjustment for contrast. In the a and b curves, there is a standard LAB adjustment with a very small movement of the center point from 'zero' away from magenta in the a curve and away from yellow in the b curve.

In this particular image, I would have dealt with the raw file in ACR by adjusting the white balance, but am doing it in this fashion to demonstrate a LAB curve adjustment.

http://gefillmore.smugmug.com/photos/119712199-S.jpghttp://gefillmore.smugmug.com/photos/119712574-S.jpg
Original ---------------------Curves applied

http://gefillmore.smugmug.com/photos/119712786-M.jpg
Curves adjustment

Contrast adjustment-

Again, as stated previously, you generally do not use LAB adjustments on bright colors. However, reading Margulis, LAB Lightness curve adjustments work well on bright colors. Let's use the bottle image again.

Here are images of the RGB channels of the bottles:

http://gefillmore.smugmug.com/photos/119713273-M.jpg
Red channel-----------------Blue channel-----------------Green channel

This is the Lightness channel in LAB:

http://gefillmore.smugmug.com/photos/119711837-M.jpg

There is a wide disparity from light to dark in the RGB channels with an image with bright colors, making it difficult to adjust contrast by steepening a curve in RGB mode. In the Lightness channel, it is a short throw. This shows that contrast adjustment is easier in the Lightness curve in the LAB mode than dealing with contrast in RGB mode in this particular type of image. But, it also means that the adjustments need to be small.

Please see the images in the color casts section above for contrast improvement through use of the Lightness curve adjustment in LAB mode.

Targeting specific colors-

The a and b curves are opponent-color channels. The colors in the a curve are magenta in the light portion and green in the dark portion. In the b curve, the colors are yellow and blue in the respective portions.

Because the a and b curve in LAB deal specifically with color and not contrast, specific colors can be adjusted without affecting the image as a whole and without affecting the other colors.

We'll use the bottle image again:

http://gefillmore.smugmug.com/photos/119712574-S.jpg


In the first example, the red/magenta/orange will be transformed to green:

http://gefillmore.smugmug.com/photos/121337865-M.jpg


The a curve is the magenta/green curve. The lower half of the curve is moved from magenta to green, transforming the red/magenta/orange to green.

In the second example, the blue/teal will be changed to red/magenta.

First, we change the blue/teal to green:

http://gefillmore.smugmug.com/photos/121337974-M.jpg

The blue half of the b curve is moved into the yellow portion changing the blue to green.

Then, the green is transformed to red/magenta:

http://gefillmore.smugmug.com/photos/121338021-M.jpg

The green portion of the curve is moved into the magenta portion, completing the transformation of the teal/blue to red/magenta.



There are other discussions of interest on LAB color in Chapter 4 of Professional Photoshop.

But for much more in-depth discussion, analysis, and examples, please see Dan Margulis’ book, Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace, and/or the discussion in the thread (http://digitalgrin.com/showthread.php?t=18203) here at Digital Grin.

[Duffy Pratt]

This is a very simple and clear write up. My only quibble is that the big power of LAB comes with steepening the A and B curves. The byzantine L curve in your barn example, in my opinion, overshadows what is going on in the A and B curves. (I like what you did with the L curve, but I think it is not the best thing for demonstration purposes).

To demonstrate what LAB does, maybe you could separate into two different curves, once for A and B, and then for the L move. That way, the reader could more clearly see what is going on with the LAB color moves.

Oh, and congratulations on using a LAB example that is neither a canyon nor a face.

Duffy

[gefillmore]

This is a very simple and clear write up. My only quibble is that the big power of LAB comes with steepening the A and B curves. The byzantine L curve in your barn example, in my opinion, overshadows what is going on in the A and B curves. (I like what you did with the L curve, but I think it is not the best thing for demonstration purposes).

To demonstrate what LAB does, maybe you could separate into two different curves, once for A and B, and then for the L move. That way, the reader could more clearly see what is going on with the LAB color moves.

Oh, and congratulations on using a LAB example that is neither a canyon nor a face.

Duffy

so you're saying to correct with a and b first, showing the barn and the a and b correction curves and then correct with the lightness curve next, showing the l curve and barn after l curve correction-

I think you have a good point but I don't want to do this then hear you say, "no, good grief no, dummy, that's not what I meant!"-

[Duffy Pratt]

Yes, good grief, yes dummy. That's exactly what I meant. :D

Duffy

[gefillmore]

Yes, good grief, yes dummy. That's exactly what I meant. :D

Duffy

thanks duff-

again, good point-

[rutt]

Really good job, George! Now I can even read the text.

I've seen this over and over again now. If you are going to invest in learning just one set of Photoshop tools, spend a weekend learning to do the basic LAB corrections described in this chapter and the first few chapters of the LAB book. It's the most bang you can get for your time investment.

[gefillmore]

thanks rutt-

andy and davidto- thanks for your help-

[Nikolai]

Nice writeup, thank you for your work! :thumb

I agree with Duffy's suggestion though.. :rolleyes

Can't wait till my turn (I'm #6 :-)

[gwilson]

This was a wonderful write-up. I have just started the book, and these discussions will be really helpful. I do have a question--in the example of your lightness adjustment on the barn I noticed the brush on the horizon became rather posterized. I suppose a fast and relatively effective way to correct that would be to simply brush it out on the layer mask. How radical can you get with the lightness curve? Could this be corrected by fiddling with it, while maintaining the dramatic clouds?

Gloria

[jfriend]

This was a wonderful write-up. I have just started the book, and these discussions will be really helpful. I do have a question--in the example of your lightness adjustment on the barn I noticed the brush on the horizon became rather posterized. I suppose a fast and relatively effective way to correct that would be to simply brush it out on the layer mask. How radical can you get with the lightness curve? Could this be corrected by fiddling with it, while maintaining the dramatic clouds?

Gloria

That's a good catch in the Lightness curve. I hadn't seen that. It's caused by the lightness curve going to a negative slope. If there are any tones that actually exist where the lightness curve goes to a negative slope (starts going down instead of up), then you get this really funny look there.

It isn't technically posterization, it's just reverse contrast - but it does look funny either way. I myself never use curves this extreme. If I find I want this much of a different adjustment for the low tones as the high tones, then I often find I need to use two curves, one for each tonal range each controlled by a mask or by blend-if settings. The two-curve techique is able to avoid these areas going all funny like you saw, but of course, it requires a valid mask or other blending setting to control it properly.

In this particular example, you could probably just block the curve from the foreground and lower part of the sky and still get the barn and sky effect you were looking for as you suggested.

[gefillmore]

This was a wonderful write-up. I have just started the book, and these discussions will be really helpful. I do have a question--in the example of your lightness adjustment on the barn I noticed the brush on the horizon became rather posterized. I suppose a fast and relatively effective way to correct that would be to simply brush it out on the layer mask. How radical can you get with the lightness curve? Could this be corrected by fiddling with it, while maintaining the dramatic clouds?

Gloria

wow! somebody's read my stuff-

re horizon- I really did not know what I needed to do about that and I suspect that john is probably right about the negative slope; you look at my example and you can see the slope-

I will try to check this out further tomorrow-

thank you very much for commenting and glad you're here on the forum-

[gefillmore]

Nice writeup, thank you for your work! :thumb

I agree with Duffy's suggestion though.. :rolleyes

Can't wait till my turn (I'm #6 :-)



thanks much, nik-

[gwilson]

It isn't technically posterization, it's just reverse contrast - but it does look funny either way. I myself never use curves this extreme. If I find I want this much of a different adjustment for the low tones as the high tones, then I often find I need to use two curves, one for each tonal range each controlled by a mask or by blend-if settings.
In this particular example, you could probably just block the curve from the foreground and lower part of the sky and still get the barn and sky effect you were looking for as you suggested.

Thanks for your response, John. It looks like this would be handled just like the RGB curves I usually work with. I wanted to rule out the possibility that the Lightness channel has magical properties that could accomplish this (efficiently) on one curve.

Gloria

[gwilson]

wow! somebody's read my stuff-

re horizon- I really did not know what I needed to do about that and I suspect that john is probably right about the negative slope; you look at my example and you can see the slope-

I will try to check this out further tomorrow-

thank you very much for commenting and glad you're here on the forum-

It looks as though about 3,000 others have 'read your stuff'! Thanks for your response and also the welcome-

Gloria

[gefillmore]

re barn background 'posterization'-

1) john says that 'I myself never use curves this extreme'. I also had that in my writeup and obviously did not follow my own advice.

2) a part of the 'posterization' and unnatural look to the background is that part of the trees has somewhat of the same 'tone' or 'shade' as the sky and that was transformed along with the sky.

3) from what I understand, posterization occurs when there is not enough 'bit depth'--can you have a section of a photo that has less 'bit depth' than the rest of the photo?-sorry, that probably doesn't make any sense but I'm trying to figure out the why of this-

probably the best way to deal with this is like john and gloria mentioned, some type of masking-

[Duffy Pratt]

re barn background 'posterization'-

3) from what I understand, posterization occurs when there is not enough 'bit depth'--can you have a section of a photo that has less 'bit depth' than the rest of the photo?-sorry, that probably doesn't make any sense but I'm trying to figure out the why of this-
Posterization occurs when you take data lying accross a given range, and you compress that range too much. For example, if you had data that previously lay along 230 to 245 in a channel, and then you steepened the curve in that area enough, that same data might be forced to fit from 235-237.

This picture of the Samoyed has been posterized in two ways. I converted to LAB. First, I created a threshold layer, set it to luminosity, slid the slider just a bit, and then reduced opacity to 63%. Then for fun, I wrote an impossibly steep curve on the B channel. See how I got lots of bands. In an example like this, bit depth has little to do with it: You could get this kind of posterization no matter how many bits you started with.

The bit depth idea for posterization is as follows: the 16 bit crew believes that writing ordinary curves on everyday pictures causes posterization. They believe this because the histogram tells them so, and because it logically follows from the math. The interesting thing is that the posterization that they are complaining about almost never actually shows up in any working example using 8 bits.

Duffy

[gefillmore]

duffy-

ok, I think I got that-

but why specifically does that little area on the barn pic so readily 'posterizes' and/or looks unnatural way before any other part of the pic does-

[gefillmore]

duffy-

you and jfriend need to do one of these chapters-

in fact, you should of done this one, although I've learned a lot from it summarizing it--

this is hard for someone who's attention deficit-

that reminds me, I've got to get back to the writing of my book, Meditations for the ADHD-

[Duffy Pratt]

I thought I agreed to do the chapter on ACR. I might be willing to do some other chapters as well.

I played some with the Barn picture. The negative slope comes largely because you were determined to put some contrast in the sky, and this basically causes you to write conflicting curves. The issue may be better handled with the Shadow/Highlights tools, or with the mind boggling stuff that Dan does in the last chapters.

For this version, I converted to LAB and wrote a simpler curve that kept the extreme steepness in the barns, but then did not go negative in slope before curving back up to hit the sky. For the A and B channels, I went even further than you had. I kept the B positive change to about 60% of the A change, but hit the B negative hard with some control points, to try to coax some blue into the sky. (This might have been a mistake).

To darken the sky highlights, and to lighten some of the shadows, I used the Shadow/Highlights tool on the L channel. Settings were pretty extreme, especially on the highlights (something like 75% amount, 50 width, and some huge radius).

Even with that, I didn't have anywhere near the drama that is in your picture. So I did a couple of blends. First I blended the B into the L channel in Overlay mode. That darkened up the barn pretty well and did some nice contrast moving in the greenery.

Then to just darken the whole thing, I applied the L onto itself in multiply mode at 40%. I didn't mess with sharpening, because it's really tough to get it right on images with this little resolution.

I won't say that I like my version any better than yours. The horizon problem is basically gone. The banding that you see in mine, if you do, comes from the natually brighter light coming from the horizon under the clouds. The trees have kept their detail, and overall I think that area of the picture looks better.

Anyway, I think its probably helpful to show what happens when you really hit the steepness pretty hard in the A and B channels. And I think it also shows that this original can stand several different types of treatment.

Duffy

[Duffy Pratt]

Oh, and just for fun I decided to darken the whole thing using another of Dan's strange steps. This is the same picture as above but with a false profile with a gamma of 3.0. I don't think any curve could get you this kind of change. And it took less than 3 seconds to do. (Just a sneak peak ahead.)

Duffy

[gefillmore]

ok, neat-

I was not sure you could do anything to the pic without affecting the trees adversely-

thanks much for your time and explanation-

re summary: I did not see your name listed-

which chapter is on ACR? (I think I left my book at work)-

[Duffy Pratt]

Chapter 16 is on ACR. I may not be on the list. Rutt asked me to do a summary when I asked some questions about Dan's ACR recommendations. At the time I thought I didn't have enough experience with ACR to do a decent job. I'm still not sure on that, but I'm willing to do the summary.

Duffy

[gefillmore]

Chapter 16 is on ACR. I may not be on the list. Rutt asked me to do a summary when I asked some questions about Dan's ACR recommendations. At the time I thought I didn't have enough experience with ACR to do a decent job. I'm still not sure on that, but I'm willing to do the summary.

Duffy

hah!-

re decent job-

I frankly don't see that as a problem-

and if they let me do chapter 4, I'll bet anything he's overlooked putting you down for chapter 16-

[Myer]

Ok,
I have a question relating to the barn in Chapter 4.

Preamble to question -
On page 59 Dane indicates the value of choice for Shadows in RGB and CMYK are 15/15/15 and 80/70/70 and for Highlights 245/245/245 and 5/2/2.

I understand that.

On page 83 he say "The L curve is a simple straight line to establish the highlight".

Maybe I missed it, but where does he indicate the values of choice for Shadows and Highlights in the L channel?

My guess would be in the low, mid 90's. Say 93 or so. But I don't remember him giving us a value.

- - - -
After going through the LAB book I thought I knew something. I found out how much more I have to do after looking at page 84. I have a lot of work to do to get comfortable with "what happens to this when you do that" or "what do you have to do to get a certain result", etc.

I hope that constant headache isn't about to come back.

[Duffy Pratt]

I haven't done it, but if it bothers you, set a highlight and shadow in RGB, and pick info points at the highlight and shadow. Now switch the info tool to display lab for those points instead of RGB. Voila, you now know what the lab equivilent of 245,245,245 and 15,15,15 (or whatever) is.

I've pretty much settled on somewhere around 92 for the highlight in LAB and anywhere from 5-10 for the shadow depending on the image.

Duffy

[Myer]

I haven't done it, but if it bothers you, set a highlight and shadow in RGB, and pick info points at the highlight and shadow. Now switch the info tool to display lab for those points instead of RGB. Voila, you now know what the lab equivilent of 245,245,245 and 15,15,15 (or whatever) is.

I've pretty much settled on somewhere around 92 for the highlight in LAB and anywhere from 5-10 for the shadow depending on the image.

Duffy

= = = =

Those sound like good enough numbers. Thanks.

M

[Duffy Pratt]

I did the check: 245s is the same as 97 in the L. 15s is the same as 4 in the L. So I've been too conservative with my LAB corrections. Of course, the L curve feels sort of like a sledgehammer compared to the lighter feel of RGB or CMYK.

Duffy

[gwilson]

I did the check: 245s is the same as 97 in the L. 15s is the same as 4 in the L. So I've been too conservative with my LAB corrections. Of course, the L curve feels sort of like a sledgehammer compared to the lighter feel of RGB or CMYK.


There is a good explanation of how Dan sets the highlight and shadow values on page 89; he says an agressive value would be 98 and 3, while a more conservative value would be 94 and 10. However, he points out that we need to "resist the temptation to overuse it", and it's "easy to fix those things in other color spaces"...L being a sledgehammer, as Duffy says.

Gloria

[Myer]

There is a good explanation of how Dan sets the highlight and shadow values on page 89; he says an agressive value would be 98 and 3, while a more conservative value would be 94 and 10. However, he points out that we need to "resist the temptation to overuse it", and it's "easy to fix those things in other color spaces"...L being a sledgehammer, as Duffy says.

Gloria

= = = =

Ok,

So 93-94 and 8-10 it is. I really don't think a bot here or there makes a big difference.


Thanks.

[Myer]

I was wondering how others did with the quiz on Page 84.

After reading the LAB book, I was quite disappointed that I started having so much trouble with this quiz.

I really learned a few things that made it easier to complete and for me to understand.

[Myer]

I was wondering how others did with the quiz on Page 84.

After reading the LAB book, I was quite disappointed that I started having so much trouble with this quiz.

I really learned a few things that made it easier to complete and for me to understand.
= = = = =

I really learned a few tricks here about cause and effect regarding interchanging LAB 'a' and 'b' channels.

This refers to the quiz on page 84.

If anybody is interested in this, I'll post the tricks that allowed me to complete the quiz almost flawlessly.

Otherwise, I won't bother.

[rutt]

Hey, man, shoot. I remember that I had to do something like that for this quiz as well. That chapter 1 quiz, on the other hand, well you have to get your head screwed on real tight.

= = = = =

I really learned a few tricks here about cause and effect regarding interchanging LAB 'a' and 'b' channels.

This refers to the quiz on page 84.

If anybody is interested in this, I'll post the tricks that allowed me to complete the quiz almost flawlessly.

Otherwise, I won't bother.

[gefillmore]

please. I'm game-

I don't mind cheating on it-

[Myer]

Hey, man, shoot. I remember that I had to do something like that for this quiz as well. That chapter 1 quiz, on the other hand, well you have to get your head screwed on real tight.
= = = = = = =

First about LAB.

Colors in the "a" channel go from solid green (very -) to neutral (white,grey,black) to solid magenta (very +).

Colors in the "b" channel go from solid blue (very -) to neutral (white,grey,black) to solid yellow (very +).

If you select a part of the photo that has a very neutral portion, it doesn't matter how you interchange the channels with each other. The result will always be neutral. Think about that for a second.

That's because each channel is 0 (zero) or very close to it and when you interchange or invert it still remains at 0 or close to it.

Therefore, since neutral gives you so many posibilities, that's of no value to you in determing the solution to the quiz.

On the other hand, if you look for a portion of the original image that is very solid green, magenta (I try to leave this out since I'm never sure what solid magenta is), blue or yellow, what you now know is that the other channel will be 0 (zero) or close to it.

For example, if part of the image is solid green, then the the "a" channel will be very - and the "b" channel will be 0 or close to it.

That means if the "b" channel or inverse of the "b" channel is used for the "a" and "b" channels, that part of the image will become neutral or 0.

Let's get practical. If part of the image is solid green to start and become neutral (white, grey or black) then you know that the original "b" channel
or it's inverse must have been used for both channels. That's because neutral in the original "b" becomes neutral in both channels. Even when inverted.

If however, the area that was originally green becomes blue, then you know the "a" channel was moved to the "b" (not inverted) and the "b" was moved to the "a" (regular or inverted).

Likewise, if the area that was originally solid green becomes solid yellow, then the "a" channel was moved to the "b" and inverted and the "b" was moved to the "a" (regular or inverted).

By selecting one other original area of interest (solid blue or solid yellow) you can nail down any variables such as when a channel could be moved regular or inverted.

One other thing I used. Cyan is green/blue and violet is magenta/blue. I used this somewhat when the result was cyan or violet.

Dan used another trick which I called "cheating" and I tried not to use it.
He kept track of the combinations used and immediately eliminated those as a future possibility.

As you get closer to the end that really makes it go faster since there can only be 4 combinations of each set of channels. If three are already used, you don't have to worry about inverted channels.

I learned the above just after I started the quiz. I would have thought I'd know this after reading the LAB book.

I hope this helps.

[dandill]

I hope this helps.For sure! Your post goes into my copy of the book. Thanks very much.

[gefillmore]

thanks myer-

hope to check it out this weekend-

although it makes my head spin-