This section covers the blending modes lighten, darken, luminosity, and color. Here are definitions of these blending modes when A is layered on top of B with the given blending mode:
- Lighten: Where A is lighter than B use pixels from A; otherwise, use pixels from B.
- Darken: Where A is darker than B use pixels from A; otherwise, use pixels from B.
- Color: Derive color values from A and luminosity from B.
- Luminosity: Derive luminosity from A and color from B.
The distinction of luminosity versus color is easiest to understand in the L.A.B. color model where luminosity has it's own channel. In L.A.B. a color blend takes the L channel from B while retaining the A and B channels from A. A luminosity blend does the opposite.
Note that in some important sense, Lighten doesn't actually make anything lighter. It merely chooses the lighter corresponding pixel from the two layers. If B is lighter than A the blend does nothing. Similarly, Darken may not actually make anything darker, it only selects the darker of the two layers. Dan Margulis remarks that really Lighten means don't darken.
Color Blend The words white and black are both neutral in terms of color. So in color blending mode, they have the effect of neutralizing the background. The word red is, well, red, so it blends by making everything it covers red.
Luminosity Blend In luminosity blending mode, the contrast comes from the upper layer. The words black and white have the effect of removing color from the blend because there can be no color when luminosity is either completely dark or white. The word red is roughly neutral in terms of brightness and so it picks up the underlying color and imparting a neutral brightness to the blend.
I use these four blending modes more than any others by far. To show why, I'm going to walk through the processing of this image. Shown at the top is the final version. Shown right is the original. In order to correct it, I'm going to break the problem into two separate subproblems: Make the best B&W conversion I can which will give me the contrast I want, and then worry about color. Well, I cheated a little. I used As Shot for Color Balance in raw conversion instead of being more more careful. This results in a blueish cast and overly magenta skin tones. In actual practice, I would have used custom color balance to make some point on his shirt neutral. But I wanted to expose both color and contrast color corrections for the purposes of this tutorial.
Shown right is the Green channel. In general, flesh details are best in the green channel, and this is no exception. Notice especially that the right side of his face (our left) is blown in the red and has nice detail in the green. On the other hand, I do like the contrast of wall and windows in the red channel. The windows are darker and the wall has more weight and detail...
Shown left is the Blue channel. Not much to recommend it, too dark in the foreground subject. If anything, the contrast in the wall and windows is worse even than in the green. Except there is much better detail under the helmet in this channel than in either of the other two. You may have to look at this image large in order to see this. But it's dramatically better.
Shown right is the default conversion that results from Image->Mode->Grayscale. Overall, it seems too dark. I'd like better detail on the subject (from the green) and better contrast on the wall and windows (from the red). And I'm greedy. I also want to unplug that area under the helmet (from the blue.)
Shown left is the red channel layered over the green channel in Darken mode. Because the man was darker everywhere in the green than in the red, he is not affected by the blend. But those windows are darker in the red than in the green and so this blend has picked up that difference. Also notice the slightly darker brickwork. Compare this to the separate green and red channels to see this.
Now what about that area under the helmet? I took the version above and layered the blue on top in Lighten mode. The result is shown right. Because man in the the blue was generally darker than the other two, it had no effect on him. The area under the helmet was indeed lighter in the blue (which was the whole idea after all) and so it is blended from the blue just as I wanted. Great. But what about that wall and windows? They are lighter in the blue than in the previous blend. So the lighten blend has picked them up. Darn.
Blend-if to the rescue. As it turns out that area under the helmet is truly plugged in the red channel, much darekr than the wall and windows. At left is the blending options dialog box for the blue layer. Notice I've moved the right hand slider on the bottom nearly all the way to the left. This has the effect of excluding from the blend all the parts of the image that aren't pretty dark in the underlying layer (in this case the red-green blend). Since the target area under the helmet is the only area which is this dark in the red-green blend and also lighter in the blue layer, the effect of the blend is limited to unplugging the hat band.
Here is the final B&W conversion. I've gotten the good flesh contrast from the green, the good wall and windows from the red, and the good helmet from the blue.
This blend has achieved something that I could never have done with curves by moving whole areas of the image into different contrast zones.
Now I have a B&W version that I like. Well, that's one way to look at the situation. In fact, I've also done something else. I've corrected the contrast of the image in isolation from the problem of correcting its color. What if I used the contrast from the B&W conversion with the color from the original? Shown at left is the original with the final B&W version layered on top in Luminosity mode. Compare with the original. The flesh detail is better, the shirt has more weight and depth, the windows are darker than the wall, and that pesky headband is unplugged.
But the color still stinks. There is that horrible blue cast, resulting (among other things) in overly magenta flesh (since blue is the opponent of yellow, and we need more yellow in the flesh.) Also there is good reason to think that shirt was black instead of dark blue. And the wall also is blue enough to be suspicious. So let's just guess that there is an overall blue cast, stronger in the shadows than in the highlights. I made curves adjustment layer on at the top of my image and darkened the blue through the midtones. This curve is shown right.
One of the annoying things about RGB and CMYK curves (as opposed to L.A.B. curves) is that they combine contrast and color. So the move above could make the image quite a bit darker and also lose contrast where in the blue midtones where the curve flattens. I don't want that, just better color balance. Color blending mode to the rescue. By changing the blending mode of the curves adjustment layer, I remove these worries. Only the color and not the contrast is impacted by the blue curve. Shown at left is the image with the corrected color balance from the curves with Color blending mode.
I wanted more color differentiation and better fleshtones. So I converted to L.A.B. and used a standard L.A.B. curve steepening move to do this. This is outside the scope of this particular tutorial, but see here for details. It getting there, en? But what about that blown spot on his left (our right) cheek? Doesn't it look odd that it's so white given how tan and sun burnt he is? I want to get a little color into that area.
I made a duplicate of the topmost layer and painted the fleshtone from the unblown part of that cheek over the blown parts. (Shown at left.) Now the color is right, but it's a mess because I've obscured the detail and contrast with my brush strokes.
Shown right is the image with the lowered opacity for the flesh tone fix layer. The difference between this and the final image shown at top is sharpening and a slight curve adjustment for the shadows. Summing up: The workflow I used for this image relied almost entirely on the four blending modes, lighten, darken, color, and luminosity. I used lighten and darken blending modes to combine the three RGB channels into a B&W conversion with the best parts of each. Then I used luminosity blending to use the contrast from this B&W conversion and the color from the original. Lastly, I used color blending mode twice. Once to allow me to color correct with RGB curves without worrying about the impact on contrast. And again to get some color into the blown fleshtones without having to worry about careful painting (which would have been way too difficult for me, given my skills as a draftsman.) This dgrin post is another demonstration of essentially the same workflow. The results are more dramatic because the original was shot in midday sun and had worse contrast problems. You can discuss this tutorial in Dgrin.