Using LAB to transform extreme colors

rutt

andy wrote:

:puke it was lime green and hideous. that window designer should be fired, but hey, what do i know about window designd :lol3

thanks john.

Try inverting the LAB A curve? That makes green magenta and visa versa. Just pull the leftmost endpoint all the way down and the rightmost one all the way up (or visa versa, whatever is the opposite of how it starts.) This often results in really surprising but realistic color shifts as long as there are no flesh tones you need to preserve.

This is a very interesting image, but it lacks punch. Color might provide it, if you can figure out how to do it without tossing your cookies. The color of the flags in your other St. Pat's day post was so great, it really made the shot.

rutt

Using LAB to transform extreme colors
Andy posted a B&W shot of a store window and reflection here. I suggested that he show the color version but he declined because "it was lime green and hideous." This set me off on a tagent about how to use LAB to have fun with extreme colors. It seems more appropriate to do this here.

First off, if you don't already understand the LAB color space, you'll need some background to understand what follows. You should look here for my introduction. Optionally, you can look here and here for a deeper explanation from Dan Margulis, a master color theorist and retoucher.

OK, if you've come this far, we are ready to have a little fun. Here is a shot of mine of a NY store window. Hope you like red.



Don't like red? How about green?



Or cyan?



OK, OK, maybe we can even find something tasteful. How about blue?



Well, maybe I'd better leave the high fashion color to the pros. Let's focus on the color theory.

Basically we can transform that bright red into any color we like with one application of LAB curves. I transformed the original red version to green with this A curve (no change L,B):



I transformed the green version to the cyan version with this B curve (or by starting with the original red version and applying both the A curve above and this B curve; it comes to the same thing.)



The "tasteful" blue version resulted from the following A+B curves applied to the original red version.



What's going on here? The red in the original version very saturated. Here are some screen shots of the A and B curves with some interesting points marked.



Notice that the bright colors in the window (I've labeled these "dress") are way outside the range of natural colors in the fleshtones, sky, building, etc. That means that if we move these points in the curves and keep the other points constant on the new curve, we can convert the bright color represented by these points into any color we like. Of course, we won't change the brightness (luminosity) by doing that; we'll have to monkey with the L curve to accomplish that. Any changes in the A+B curves only effect color, not brightness.

Look at the A curve I used to make the green dress. This is almost a pure inverse curve, mapping green to magenta and magenta to green. It isn't quite a pure inverse, which would have mapped the somewhat magenta fleshtone of the faces to an unattractive slightly green tone. So I set a couple of points in this area of the curve before pulling the left endpoint down and the right endpoint up. Now it maps colors outside this range to their inverses, but leaves the less saturated colors alone.

This is all a lot easier to do in LAB because color tone is strictly separate from brightness. Very saturated colors like the one in this window rarely occur in nature, certainly not in fleshtones. So often LAB gives us the freedom to change these colors in any way we like without any change to the naturally occurring colors. Of course, there are exceptions to this -- flowers, grass, blue sky. But often these are not an issue. In the window shot, there is very blue sky, but it's on the other side of the B curve from the yellow in the dress, so we can retain it and still hit the dress.

So how about it, Andy? Let's fix that lime green in your window and get a great color "Silent Watcher" image.

DavidTO

Very cool.

Dan's book is sitting in my living room. I've gotten through 20 pages(or whatever, not sure, actually, but it feels like 20 pates!). Busy life, ya know? Wish I could just download it without all the work!

rutt

I'm not sure you'd get this particular thing from Dan's book. I took Dan's 2 day cram course. When he introduced LAB, Dan inverted an A curve to make a red car green. Sort of a magic trick. But Dan really isn't about this kind of thing; his goal is to make great realistic looking colors, to try to make the image be what the photographer really was trying to capture. The kind of games I'm playing here are really photographer's games, reserved for those with the creative right to completely tranform the image.

BTW, Dan's 2 day course is about as close to a download of his book as you can get. It's 2x12 hour days. Your head explodes. But it's a great nerd out and in the end you have about 40% of the book cold and the rest is much more accessable. I met some really interesting people. Someday, I'll take the 2 day advanced course. Then nobody will be able to understand a word I say.

DavidTO

rutt wrote:

I'm not sure you'd get this particular thing from Dan's book. I took Dan's 2 day cram course. When he introduced LAB, Dan inverted an A curve to make a red car green. Sort of a magic trick. But Dan really isn't about this kind of thing; his goal is to make great realistic looking colors, to try to make the image be what the photographer really was trying to capture. The kind of games I'm playing here are really photographer's games, reserved for those with the creative right to completely tranform the image.

BTW, Dan's 2 day course is about as close to a download of his book as you can get. It's 2x12 hour days. Your head explodes. But it's a great nerd out and in the end you have about 40% of the book cold and the rest is much more accessable. I met some really interesting people. Someday, I'll take the 2 day advanced course. Then nobody will be able to understand a word I say.

Huh?

Baldy

That's really cool, Rutt. I've read maybe 6 Photoshop books and I just keep going back to Dan's. I had my doubts about it early on, but no more.

pathfinder

John, I need you to keep hitting away on the utility of LAB mode. I shot this image of pelicans in Bosque - the compostion was OK, but the color seemed to lack punch. I captured the reflections but they were not dramatic for the viewer.



I reopened the psd file and shifted to lab and played with the A,B curves and burned in the midtiones in the Lightness channel in the watery foreground, and created this newer version, which I find improved. Yes, no??

rutt

pathfinder wrote:

I reopened the psd file and shifted to lab and played with the A,B curves and burned in the midtiones in the Lightness channel in the watery foreground, and created this newer version, which I find improved. Yes, no??

Nice shot and much better. I think you can do even better, though. I'd like to see better feather detail on the body of the left hand bird. There are two ways to go about this. Steepen the very highest highlights of the L curve (just like snow). Or make a black channel with good detail and blend it into the existing one in CMYK (see DM, "Plate Blending as Poetry".)

pathfinder

rutt wrote:

Nice shot and much better. I think you can do even better, though. I'd like to see better feather detail on the body of the left hand bird. There are two ways to go about this. Steepen the very highest highlights of the L curve (just like snow). Or make a black channel with good detail and blend it into the existing one in CMYK (see DM, "Plate Blending as Poetry".)

I took the image I had darkened and sharpened in LAB and steepened the lightness curve like you described.

My first impression on the monitor image was that the birds had turned ever so slightly grey and I did not think it was an imrovement. But when I printed the image on a nice white paper and examined the prints side by side under my Ott light, there is more detail in the feathers of the bird on the left and they do not look grey under a nice Ott lite side by side with the first image.

This is the starting image


And this is the image with the steepened L curve that I printed for comparison.


Another alternative is to select the whiter area with selective color and duplicate the selected areas and multiply blend it yielding this image

rutt

There are lots of ways to work on this. I thought of another thing to try. shadow/highlight, but use the highlight instead of the shadow to bring up the detail in the highlights. You might even do blend the result of doing this into the original with multiply-blend-if and < 100% opacity.

adrian_k

kewl
hey Rutt, this is excellent, gotta give this a try. I always forget you can mark certain objects on the curve. thanks.

A simpler (and therefore less flexible) method is to use a 'selective color' adjustment layer. Then you can also paint with black to ignore (mask) certain areas that you don't want changed.

rutt wrote:

Andy posted a B&W shot of a store window and reflection here.

Notice that the bright colors in the window (I've labeled these "dress") are

DoctorIt

With the window red/green/blue trick - are you really manipulating the whole image? or selecting only the window in a seperate layer or something? Cause I look at the "blue" skay in the background on the right there, and it looks about the same in each case? I guess I kinda don't get it

rutt

DoctorIt wrote:

With the window red/green/blue trick - are you really manipulating the whole image? or selecting only the window in a seperate layer or something? Cause I look at the "blue" skay in the background on the right there, and it looks about the same in each case? I guess I kinda don't get it

That's sort of the point. All the transformations I did were global (no selections, no layers.) But I controlled the colors I transformed with the curves. In the case of the blue sky background, it really is very blue. That means that the A channel is neutral and the B channel is positive.



So if I keep the 73,73 point constant in B curves I write, the sky color won't change. I can do anything I want to the A curve so long as the middle of the curve stays close to 0,0 because it that's where the sky lives on the A curve. So go back and take a look at all the curves I posted. In all cases, I anchored the B curve near the 73,73 point to keep the sky from changing and kept the center of the A curve close enough to the center to avoid changing the sky (and flesh.)

I hope that helps.

adrian_k

you're right
It's clearer to understand using the Selective color layer because you do select red/yellows/greens/etc. So you are only adjusting a certain range of colours.

And with the shop window example there are definite blocks of individual colours so it gives a marked change.

Lab mode works in a similar way but I don't understand it well enought to explain.

DoctorIt wrote:

With the window red/green/blue trick - are you really manipulating the whole image? or selecting only the window in a seperate layer or something? Cause I look at the "blue" skay in the background on the right there, and it looks about the same in each case? I guess I kinda don't get it

rutt

adrian_k wrote:

Lab mode works in a similar way but I don't understand it well enought to explain.

LAB curves and sharpening are my very favorite PS color correction tools these days. There are still reasons for being able to work in CMYK and RGB, but I go to LAB first usually. So if LAB is still a mystery, I hope you will go back and do at least some of the assigned reading from my first post, especially http://www.dgrin.com/showthread.php?t=2042
I think you'll be glad if you get a handle on what's going on here.