Lab color adjustment -- am I getting ahead of myself?

Plasmodium

Hi all,
I've been playing quite a bit with color adjustment in Lab mode. I'm not sure if this falls in a future chapter, but I've been using the a and b channel much like I use Lightness. In other words, rather than just altering the endpoints / slope, I'm alt-clicking to put intermediate points in. The trick seems to be that many of the midtone points are very very close together, so adjusting the curves can be very tricky.

Here's an example. I took this picture last month of the Petergof palace in St. Petersburg, Russia. The lighting was very challenging at first (thankfully the sun came out and I have some real gems from there too). The first is the RAW image. The second is with the Lab curve adjustments as shown, in which I only alter the slope of the a and b curve. But I wanted to get rid of the green cast on the clouds (I sometimes get that with my wide angle lens), enhance the yellow of the buildings and statue, enhance the green of the grass, soften the blue in the spruce trees yet add a bit to the water. I didn't feel like I could satisfactorily do it by altering the slope alone.

RAW image:


#1


Curves for #1


#2


Curves for #2

Plasmodium

No one has responded after a few days -- maybe I should reframe the question.

There are situations in which I want to make regional rather than global color adjustments to an image. I've really taken to using Lab channels to modify global color casts. But, there are pictures in which I want to increase green in some areas decrease it in others, and to varying degrees, all without causing a global green or magenta color cast. The problem is the Lab color channels are not very conducive to ctrl-clicking on various places on the image and moving them independently (the way you can in the Lightness channel). The points end up falling very close together on the curve. So when I have to make regional color adjustments I still find myself stuck going back to RGB where the pixels really are distributed linearly over the curve, and I can really get 5 or 10 independent points to move around on the curve. What's the solution in Lab?

DavidTO

I didn't see this until just now.

Yeah, basically you're getting into Chap. 4 and beyond. You can do those types of adjustments in a+b, you just have to understand what you're doing first with the basic recipe.

Do you have the book that we're using in the discussion group? The Chap. 4 summary should be up soon.

DavidTO

Looky here.

Plasmodium

Thanks, David. No, I don't have the book yet, but I've thoroughly read the posts and done a lot of playing around with it. The trickiest part for me technically is how close the points are to one another on the curve. I'll look forward to the chapter 4 summary.

DavidTO

Plasmodium wrote:

Thanks, David. No, I don't have the book yet, but I've thoroughly read the posts and done a lot of playing around with it. The trickiest part for me technically is how close the points are to one another on the curve. I'll look forward to the chapter 4 summary.

I know! I wish you could blow in on the curves. It can be frustrating. You can toggle to a larger size, did you know that? It's the button at bottom right.

DavidTO

Oh, and at what point are you going to admit that I was right that you could effect one color on a curve without effecting the other?

rutt

I'm glad you are catching the LAB bug. Fun isn't it? That final version of you image really does pop.

As for your question about using RGB curves in some cases, as I said before, every image has 10 channels. The good retoucher doesn't subscribe to any religiion that one colorspace is better than another. All have their uses. If I can get better results on a particular image in RGB than in LAB, I'm there. My images often make a trip through CMYK so I can manipulate the black channel independently. CMYK is very like RGB, actually; much more so than either is like LAB.

I also said this before, LAB is the Big Bertha, the long range cannon, in your color corretion golf bag. It will get you on the green 90% of the time. Then you might need the sand wedges and putters which are available in the other colorspaces. You can convert your image back and forth from LAB to RGB all day and you won't degrade it. Learn to use LAB for what it's good at and don't lose your RGB skills. In the end you'll be a stronger post processor than if you limit yourself to just one colorspace.

I'll get down off the pulpit now. I think your particular question was this: how to change only very specific sections of the A and/or B curves. This really is a Ch. 4 question, but you are on your way to reinventing some of Ch. 4 yourself. (Hint: buy the *$&#@ book already.) And let's just suppose for a minute that a trip to RGB isn't going to solve the problem.

1. Use curves twice. First steepen the curves symetrically and equally. If the colors were so close to begin with, perhaps you won't do any damage by doing this. It probably will even be good. After you do this, you can recurve. Now the colors will be farther apart and the curve easier to work with. Keep a layer from before the steepening. You can always blend to split the difference.
2. Use lock points on the curve for parts of the image you don't want to change. Cmd-click here and there on the image to set points on the curve. Then you can move other points without affecting the locked down points.

I've looked ahead in the book even a little farther. Boy are we going to be having some fun and adding some great new tools to our color correction tool kit. Dan doesn't limit himself to curves or to a single color space. Special images require special techniques, and Dan is a master of plate blending to target very specific issues very precisely. I'm not ready to explain this stuff (I'm hoping to learn to use a lot of it through this group.)

If you do get the book, how about raising your hand to write a chapter summary?