When Too Much Is Just Right: A Decade of Changes in LAB Color Technique
A decade ago, Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace burst onto the scene. Its signature move was to enhance color—and color variation—with steep curves applied to the A and B channels. This approach worked wonders with images featuring naturally dull colors, such as the canyons referred to the book's title. Let's look at an example of this basic technique, as taught in the first edition, and then discuss how we can do better a decade later.
Simple, Fast, Effective Color Adjustment: How an Excellent Technique Was Improved and Automated
The original canyon image shown in Figure 1 is transformed into the version in Figure 2 by applying the curves in Figure 3, in LAB mode. The red tickmarks in the contrast-only L channel show the range of most of the canyon; detail is enhanced by steepening this part of the curve. The color-only A and B curves are also steepened, but they pass through their original center points, which is where they define neutrality. Therefore, color becomes more intense, but grays don't change. If you think the image still lacks color, make the A and B curves even steeper; if the color is too intense, flatten the curves slightly, always holding the center point constant.
Figure 1 An original photo of a canyon in a Canadian national park.
Figure 2 A corrected version using techniques shown in the first edition.
This color adjustment was simple, fast, and very effective for its time—but I no longer work that way. The reasons are part technological, part more experience, and part more common sense.
How vertical to make the A and B curves depends on the image. The better-known canyon in Figure 4 starts out much duller than the one in Figure 1. Yet very steep curves can transform it into the result in Figure 5. I think it's already too colorful—yet LAB would permit making it even more vivid, if desired. And why not? If done on an adjustment layer, it's easy enough to reduce opacity to taste—or to find a more agreeable way to cut back the color.
Figure 4 The grandest and most iconic canyon in the world.
Figure 5 The workflow advocated in the new edition always shoots for excessive color at first, followed by an intelligent decision on how to back off. Here, although the entire image is too colorful, the blue background is more objectionable than the red rocks.
I don't accept either Figure 4 or Figure 5, but Figure 5 tells me more about what's possible with this image than Figure 4 does. To me, the blue background is more objectionable than the red foreground. Therefore, in toning down the overall color, I would choose a method that discriminates against those blues—or possibly against all dark colors. LAB has more options for doing this than you'll find in any other colorspace.
My policy today is always to go for color overkill and then back off, rather than starting with a duller image and trying to force the right amount of color into it. As this statement implies, I no longer make custom curves. The original canyon of Figure 1 needs less color added than Figure 4 does, but why worry? They can both use the same set of curves, and then we can figure out how far to back off the color in each image.
If the same set of curves will be applied to each image, automation is the way to go. A simple action will do the trick.
Something similar to the small move between the canyons of Figure 1 and Figure 2 can be done without LAB, but something as strong as between Figure 4 and Figure 5 probably can't be. Everything you're about to see definitely can't be.
Driving Colors Apart
Now, another take on the Grand Canyon image. Based on a technique shown in the first edition of Photoshop LAB Color, this approach drives colors apart, as opposed to intensifying them generally. Back then, I needed complicated curving. This time, it's an action. Getting from Figure 6 (complete with selection) to Figure 7 is just a single click. The selection is used only to suggest to Photoshop what the important ranges of the image look like. It's not actually used to correct the file.
Figure 6 A new take on the Grand Canyon image starts with an "advisory" selection that gives Photoshop an idea of which colors are most important.
Figure 7 The new MMM action uses the advisory selection to add variation—not simply more color—to the original image.
You've probably never seen an action with this kind of "advisory" selection before. And the complexity of developing a strategy from it suggests a very long action of close to 100 individual steps. Such a monster would have been out of the question 10 years ago because it would have taken too long to run. And anyway, I didn't know enough to write it. Today, it runs quickly, and is part of a suite of such actions you can download for free.
Its advantages are evident here: The reddish rocks are emphasized, but, unlike in Figure 5, the blue background is not. We can tone down the result in many ways, or unite it with a similarly toned-down version of Figure 5. In fact, a second free action lets you combine three different approaches, in whatever proportion or intensity you like. Let's have a look, using a new image, to see how LAB opens creative possibilities that are available nowhere else.
Figure 8, the original photo, shows the rough selection that will guide the process. Or, rather, the two selections. Any Photoshop method of selecting works with this action, including multiple selections—even Select All. The result can be divided into three parts.
The version in Figure 9 is contrast-only, or in LAB terms, exclusively in the L channel. It seeks to enhance detail in areas similar to those of the selection. It's not easy to see in this image, but you should be able to detect that the lighter greens in the background of Figure 9 break away from the darker trees better than they do in Figure 8. Similarly, there's more texture in the animal's coat.
Figure 8 Any selection method, including multiple selections or even Select All, can guide the action.
Figure 9 These images break down the individual elements of the MMM action when applied to Figure 8. Here, the action adds contrast without changing color.
Figure 10 is the equivalent of the earlier Figure 7. It's not more colorful as such; instead, the colors are driven apart from one another. Figure 11 is the one that intensifies color. The major difference: The Figure 10 approach considers it acceptable to modify hue. Note that the background foliage is bluer in Figure 10 than in Figure 11.
Figure 10 This is the color-variation contribution.
Figure 11 This version adds color only.
All three of these versions merge together, reinforcing each other, to create Figure 12. Too much? Certainly, but what do you expect from a maneuver that takes no account of the nature of the image? The question now is how to back off. LAB offers infinite options. We could simply reduce the opacity of the entire effect. Or we could reduce, increase, or eliminate any of its three components—even selectively, and with much more flexibility than is available in other colorspaces.
Figure 12 Figure 9-Figure 11 combine to make this overly colored version, which reveals that the oranges are more offensive than the greens. LAB makes it easy to reduce one more than the other.
It's too easy to look at Figure 12 and say "way too colorful." Instead, look at the individual components. The background is too intense, but not by much, and many people would accept it as is. Nothing is wrong with the dark parts of the bison, but the lighter areas of the coat and parts of the tree trunks are definitely too orange. We should look for a way to tone down these parts more than the others. Here again, LAB has a decisive advantage.
Masking and Blend If
We often need to isolate objects from their backgrounds or from each other, using masks or similar techniques. It's always helpful if a certain channel has at least the beginnings of such a mask, so we don't have to create it manually. The A and B channels of LAB, which are color-only, are much more likely to qualify than any RGB or CMYK channel, all of which portray darkness as well as color. Many more objects can be defined uniquely by their color than by their darkness. That factor makes the Blend If sliders, found in the Layer Style menu of the Layers palette, far more useful in LAB than elsewhere. We could use it fruitfully in the bison image, but for clarity I'll use the flower image in Figure 13.
Figure 13 It's much easier to distinguish objects by their color than by their darkness.
To show how LAB can isolate these flowers, for the sake of argument we'll say that the picture should be transformed so that the flowers retain their color but everything else becomes black-and-white, as shown in Figure 14.
Figure 14 To illustrate how easily LAB can isolate the flowers, everything else has been desaturated.
The first part of the exercise is easy enough, whether in LAB or, say, RGB. Duplicate the base layer, and then use Image: Adjustments>Desaturate. Now the top layer is a full black-and-white. The problem is to restore the flowers, which requires some kind of layer mask or Blend If to disable the parts of the top layer that contain flowers.
In RGB, this is easier said than done. In many places, the flowers are light in the red channel while dark in the other two; this is what makes them overall red or pink. Unfortunately, the trellis in the background, which is not red but also not exactly white, is just as light as the flowers, and many reflections off the green leaves are just as dark as the darkest areas of the flowers in the red channel. So the red channel can be the start of a mask, but will need work to finalize it.
In LAB, there's no problem at all. The A channel runs from magenta to green. By excluding the top layer wherever the underlying layer is significantly magenta (see Figure 15), we can detach all the variously colored flowers from the background.
Figure 15 The Blend If command is much more powerful in LAB than in other colorspaces. To create Figure 14, a duplicate layer was desaturated, and then Blend If excluded parts of it with a magenta component.
We use a similar procedure in the B channel, which is yellow versus blue. The sky, for example, is frequently the only blue object in an image—easily isolated with Blend If.
You might think that the B channel would be able to exclude the objectionable orange parts of the bison and trees back in Figure 12, but it's not that easy. True, these areas are much more yellow than they are blue, but so is the background vegetation, so Blend If can't distinguish the two. The areas are also more yellow than they are magenta; or, in LAB terms, they're more B-positive than A-positive. However, they are A-positive, and the background, being more green than it is magenta, is B-negative. Therefore, Blend If can be used in the A channel to isolate the orange areas. They can be restored to their original dull color, or to anything in between the colors in Figure 12 and Figure 8. And this is in addition to all the options suggested by Figure 9, Figure 10 , and Figure 11.
This unique ability to find definitions of objects in the A and B channels is much greater than was the case a decade ago, when Photoshop LAB Color was published. That's why I devote so much of the Second Edition to channel structure.
I hope we've established that LAB offers creative flexibility that can't be found elsewhere. It also can create wildly excessive colors. Why would you want to do that? Well, the bison image is a good example. Figure 8 is the original. It surely needs more intensity. But can you tell by looking at Figure 8 that the bison's coat and the tree trunks will be the limiting factors as to how much intensity you can add?
The title of this article is derived from the title of Chapter 4 in the new edition: "Too Much Is Just Right." Take that principle to heart: It's easier to work with too much color than not enough—especially if you know your LAB.