Photoshop LAB Discussion: Chapter 12

edgeworkedgework Major grinsRegistered Users Posts: 257 Major grins
edited March 30, 2007 in Finishing School
[size=+2]Command, Click, Control[/size]
or
[size=+1]Meet the Man From Mars[/size]

[OPINION]This chapter highlights Dan's innovative brilliance, but also his blind spot. Brilliance, because once you're comfortable with the concepts here, they'll remain an essential part of your color correction arsenal for all time. Blind spot, because it showcases his ongoing aversion to making local corrections, so much so that, while the utility of his techniques is never in doubt, one can't help thinking, from time to time, "Yeah, but a couple of masks would be easy enough to come up with, and the result would be the same. Without the trial and error."

More and more, as we move further into the esoterica of LAB, the various techniques that Dan describes make less and less sense as isolated procedures, but, rather, need to be seen in the context of an overall workflow that involves bits and pieces of a wide variety of approaches, some of which, in the real world, involve cutting masks and making the stray local correction or two. While Dan makes some effort to establish this context, here and elsewhere in the book, it's really the kind of thing that each person has to evolve on their own. His sample images might be custom-tailored to demonstrate this or that technique, but seldom are we so lucky in a live work flow. [/OPINION]

Note: as I am not a photographer, I have no library of images from which to draw for examples. Any volunteers? At the end of my summary I will spell out Dan's thoughts on what type of image is best suited for these techniques, and, should samples be forthcoming, we can kick the ideas around.

There are two differrent techniques in this chapter, the first being the one that has generated the most interest, and which is the easiest to master. In fact, he devised it as a recipe for people who know nothing about LAB and have little desire to learn. As such, it's a one-stop shopping approach that anyone willing to delve more deeply into the topic (as is the case with all here, of course) will customize and embellish as their own needs and styles dictate. It's really a continuation, or culmination, of the processes he described in the opening chapters: using curves to drive colors apart from each other, and enhance them in ways that are impossible in RGB or CMYK.

The first sample image is found on page 244, a shot of canals and buildings in Venice, and it exemplifies the perfect subject for this treatment. Nothing's particularly wrong with the picture, it's just dead. Not enough contrast, and colors that don't do much at all. The idea here is to create a grossly exaggerated set of curves that will blast the dead tones out into the four color poles—red, blue, green and yellow—along with a sharp contrast move to the lightness channel, then lower the opacity of this curve to a point where the whole thing looks believable. Just better. For this reason, images that already have a full range of color, or areas of color brilliance, either don't require this technique, or will be thrown out of whack by it.

The recipe is as follows: in the Lightness channel Command/Control Click on an area that is suggestive of the color regions that need enhancing, to set an anchor point, hereafter called the pivot. In practice, and in the sample curve Dan supplies, this will usually be near the midpoint of the curve. Then pull the bottom anchor half the horizontal distance towards the position of this pivot point.

In the A and B channels, do something similar. Command/Control Click on an area of drab color. By definition this will place the pivot point near the center, as we're talking about dull, desaturated colors to begin with. Again, pull the bottom anchor to the right, only this time continue until the top half of the curve flattens out along the top edge. How far to pull the bottom anchor point is variable, depending on the image and the effect desired. Do this with both channels. Lower opacity to anywhere between 5% and 25% (totally subjective; set to taste), and that's it. He calls this "The Man From Mars" technique, derived from the gentleman from page 20, where we first see this approach in action. (Dan likes to duplicate the base image, apply his curve to the duped layer and then set opacity to the modified image. You get the same result by simply leaving the curve as an adjustment layer and modifying the opacity there. And you keep the curves live, should further modification be necessary. You can also use blending sliders directly on the adjustment layer, in precisely the same ways that have already been discussed.)

I found Dan's discussion on the correct setting of pivot points somewhat unclear, due to the fact that this is a far less scientific process than simply pulling anchors an equal distance and keeping the center point intact. We're abandoning all pretense of remaining faithful to the art, instead striking out on our own for uncharted regions. The "correct" result is as much a matter of choice and opinion as proper use of the technique. But if you keep the goal in focus, which is to bring out the full range of colors in the piece, you should be able to get the desired result with any image. Depending on where you initially set your pivot point, you may need to tweak it to achieve the effect. It's possible that your initial pivot for your A channel, for example, might turn the image overwhelmingly green. It's easy to imagine then pushing your pivot several notches to the right, which would turn everything overwhelmingly red. Somewhere between there is a point where both greens and reds co-exist. That's where you want the pivot to reside. Likewise for the B channel. Neither too much yellow, nor too much blue, but a healthy balance between. Small nudges of the pivot points left or right can result in serious shifts of hue, due to the extreme steepness of the curve, so be judicious. You'll know when it's right, and there is no set formula that will fit all situations.

Another thought that occurs to me with this recipe approach: it's fine if you're not really interested in the dynamics of LAB, but it seems to overlook the most powerful advantage of using the color space in the first place, which is the ability to treat contrast and color as totally separate, unrelated operations. Two different goals are in place and treating them with one set of curves, in particular when the intention is to reduce the opacity drastically, seems to hamper the efficiency of the contrast move to the lightness channel. I've found it far more effective to use a second curve adjustment layer for the lightness channel, treating it like I would any image needing contrast, rather than trying to exaggerate the move, anticipating a later reduction in opacity.

Also, as Dan points out, the same result can be achieved with the earlier technique of pulling both the anchors toward the center without a pivot point at all, simply, in this case, making the moves much more drastic. In any event, the desired goal is that neon monstrosity from page 20, or the day-glo image on 244, Figure 12.1B, with all colors in evidence. Whatever curve shapes get the job done is of secondary importance. Keep in mind, too, the possibility is very likely that you will be ignoring the integrity of the center point which may force a cast into your highlights. Toying with the blending sliders, as has already been discussed, can eliminate this problem.

________________________________

The second technique is more complicated, both to achieve, and to explain. Whereas The Man From Mars can be created with either no pivots or 1 pivot, this next technique involves setting at least two points, which will then be moved apart from each other in order to create greater separation in the colors. In theory, there is nothing new here. In all color spaces, any action to a curve that steepens the angle will exaggerate differences. In CMYK and RGB this tends to be exaggerations of contrast, stretching out tonal ranges of individual channels between light and dark. In the A and B lab channels, the desired shift will be between two areas of similar color that should be more distinct from each other.

The blue jacket on page 246 needs to have the material defining the shoulder and waist areas stand out more from the rest of the jacket. In the original, they are both very close in hue and also in value. While the image is not included on the CD, one suspects that the Red or Cyan channels would show scarcely enough difference in the areas to make a curve useful. Since the jacket is blue, the important moves will be to the B channel. Checking the numbers shows that the shoulder/waist material reads -25 while the rest of the jacket reads -45. Command/Control clicking on each of those areas gives two anchor points that can then be moved farther apart. A third area on the jacket, the inner lining, is neutral and should stay that way, so a third anchor point is placed for that part of the jacket. This point is near the 0 mark, as would be expected for a neutral tone.

While I can't post the image, these color blocks pretty much approximate the colors in the jacket. The adjustment calls for pushing the top point up, the middle point down, and leaving the neutral point where it is. Note how this causes the bottom part of the curve, the "yellow" part, to distort accordingly. Were there any yellow elements in the image, this would require additional points on that end of the curve to lock them down. As there is no yellow, it doesn't matter what the yellow curve does. Of more concern is the possibility that the modifications in the blue tones will offset the midpoint, turning the white background to some kind of imaginary color, as was described in Chapter 8. Using the blending slider to eliminate any areas from the underlying layer that are white will solve this problem handily. The A channel can receive a similar treatment, Command/Control clicking on the three separate areas of the jacket, leaving the neutral point where it is, and pulling the other two points farther apart. Since the tonal variations are so slight, the L channel simply gets an overall lightening in the quartertones and a slight darkening in the three-quartertone increasing the overall contrast. IMPORTANT: moving color anchors apart should be done in vertical directions only, either up or down. Left/right movement can create unexpected and unwanted results. Using curves that are virtually identical to the ones in Dan's book, the "before" blocks on top changed to the "after" set on the bottom.

blue_demo.gif

The second example, the triggerfish on page 249, shows what can happen when two areas of similar color compete for attention, but one is more important than the other. The yellow coloring around the fish's mouth is a characteristic trait of this fish, but the yellow hued background reduces its impact, and, in fact, there seems to be a yellow cast throughout the entire image. Even though Dan's sample curve shows only three anchor points, there are five separate areas under consideration: the yellow mask around tthe mouth, the pale yellow band just below the eyes, the yellow pattern along the back, the yellow background, and the white blotches along the belly. The brown isn't really neutral, but neither do we want it taking on a blue cast, which it could easily do if we start playing around with the yellow half of the B curve. So it's not a bad idea to lock down the blue half of the curve. Command/Control Click on each of the areas in question, with a neutral point to cover the white belly spots. The mask point and the background each moved the greatest distance, in opposite directions, the mask becoming more brilliant, the background turning more neutral. The two secondary points each moved away from each other as well, though not as great a distance. Even though I added two extra anchor points, the shape of the curve is roughly the same as the curve in Dan's book and would have been the same had I only clicked on the mouth mask and the background. The extra points are just to show how the various areas move.

There's nothing for the A curve to do here, so we leave it alone, and once again give the Luminosity curve a standard contrast move. The difference in the image after applying these curves is unmistakable, and dramatic.

fish.gif

I'm not going to deconstruct the remaining images as Dan does a better job. They all deal with the problem of treating a specific color, without damaging the rest of the image, and all use the same theoretical approach. Worth mentioning is the image of the lady with the horse, which shows that, once we enter this creative realm, there isn't always a specific correct result. There are any number of directions some images can go, and often it comes down to a personal preference rather than an ability to say "This is right and that is wrong."

The other image worth noting is the man and woman. At first glance it would appear that the entire image needs help, but in fact, her pink sweater becomes far too brilliant with a normal curve enhancement and loses detail. In this case, the green, yellow and blue anchors all pull in towards the center, as would be expected, but the magenta anchor moves in the other direction, traveling up instead of in. This pushes the pinks towards neutral, although it's doubtful that any pixels are affected by the actual anchor-point move, as the values at the extremes of the curve are outside any useful gamut. But it serves as a brace against the visible part of the curve moving in towards the center and, hence, turning more brilliant than is called for. So the precise opposite effect is achieved, through an application of the same theoretical approach, targeting one specific area for different treatment than the rest of the image. In this case, the targeting serves to shield the area in question from the effects that transform the rest of the image.

To recap: The Man From Mars technique is really just an extension of the curve moves that were introduced in the early chapters of the book. It's purpose is to enliven the entire image and assumes that no one part is more important than others. Faces, city scenes, landscapes are each good candidates for this approach. But in situations that require a greater variation in specific colors than in others, targeting specific parts of the relevant curve and driving the tones farther apart is called for.
There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
—Korzybski

Comments

  • ruttrutt Cave canem! Registered Users Posts: 6,511 Major grins
    edited November 21, 2005
    edgework wrote:

    More and more, as we move further into the esoterica of LAB, the various techniques that Dan describes make less and less sense as isolated procedures, but, rather, need to be seen in the context of an overall workflow that involves bits and pieces of a wide variety of approaches, some of which, in the real world, involve cutting masks and making the stray local correction or two. While Dan makes some effort to establish this context, here and elsewhere in the book, it's really the kind of thing that each person has to evolve on their own. His sample images might be custom-tailored to demonstrate this or that technique, but seldom are we so lucky in a live work flow.

    I had the chance to watch him in the wild when I took his advanced color theory class. He blends channels with frightening virtuosity, but he really isn't afraid to add a layer mask and make an easy selection to exclude large parts of the image. What he won't do is draw a selection where he has to get down to the pixel level to get it right.

    It's also very true that all his techniques are cogs in some giant workflow machine. The portrait technique makes no sense without the A/B channel masking techniques and the "impossible retouch" makes the most sense as part of the portrait workflow.
    If not now, when?
  • ruttrutt Cave canem! Registered Users Posts: 6,511 Major grins
    edited November 21, 2005
    Here's an example of the Man from Mars technique in action. Interesting that this is also an underwater scene.

    Original:

    45456598-L.jpg

    These curves on a duplicate layer:

    45483394-S.jpg45483368-S.jpg

    Made the jellyfish look even more like an alien than it did before:

    45456341-L.jpg

    But then I reduced the opacity of the duplicate layer to about 40%, flattened it, and applied pretty steep L curve to establish some contrast to go with the new saturation and got this:

    45456864-L.jpg

    Writing the curves for the MFMs techiques it really fun and easy. Pick a point you want to keep it's color. The further to the right it is, the cooler your image will stay in that channel. Then pull the leftmost end point inward until the curve distorts on the other side and starts hugging the top until it makes a radical dive.
    If not now, when?
  • edgeworkedgework Major grins Registered Users Posts: 257 Major grins
    edited November 24, 2005
    Love the color move. The luminosity move has some problems. It's always a trick, when you increase contrast in part of the image, to figure out where to hide the flat pixels. Since the lightest area in the image reads 53, there is clearly sufficient room at the bottom of the curve to crunch it, since nothing's there anyway. But a simple "S" move doesn't really work. significent detail exists in a number of places, all of which reside in their own value range. There are the subtle filaments feathering off the trailing rootlike forms, light at the ends, dark close to the body of the beast; there is the leafy stem with lots of complex lighting going on; the cap should look translucent: there's stuff going on inside and you've captured it just above the rim, but the top has plugged up; and there are the ribbed sections of the rim. Chances are, any effort to pinpoint a steepening in one of these areas will force flat pixels into another, equally important area. So it takes some trial and error, and trading off. This is the luminosity curve that I came up with:

    jellyfishcontrast.gif

    With this result:

    jellyfish.jpg

    I like parts of yours better, particularly the tendrils against the lighter part of the water at the left. This might be an occasion for a large soft mask to allow two different curves to work.
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • edgeworkedgework Major grins Registered Users Posts: 257 Major grins
    edited November 25, 2005
    Here's a comp using the best contrast parts from each image.

    jellycombo.jpg

    Sometimes, you just have to patch it all together in the editing room.
    There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
    —Korzybski
  • ruttrutt Cave canem! Registered Users Posts: 6,511 Major grins
    edited March 30, 2007
    Hey, Edgework, the images from your initial post seem to have vanished? Do you still have them somewhere? We'd like to preserve them somewhere permanent for posterity.

    Give me a tingle. Thanks.
    If not now, when?
Sign In or Register to comment.