The Artificial Tanned Look
Just as we remember—or desire—forests to be greener than a camera sees them, so do we adjust our perception of the human face.
Every face contains what our society perceives as defects: wrinkles, blemishes, scars, and whatnot. So ingrained is our instinct to remove these in Photoshop that we often nowadays see pictures of 60-year-olds whose skin has been manipulated to be as smooth as that on the centerfold of Playboy. Without endorsing such an atrocity, one can agree that something should be done to tone down, if not eliminate, obvious aesthetic problems in a face.
Having voted in favor of revising reality by retouching, it's hard to make a case against doing it with color as well. We visualize people as being healthy; we associate health, perhaps, with outdoor activity; but for whatever reason we tend not to like overly pale or pink skin—even if that's the kind of skin the model has. The powerful preference for suntanned skin was first noted in print at least as early as 1951, by a Kodak scientist, David MacAdam. My own tests not only confirm it but go further: we tend to consider the suntanned skin more accurate. Some ten years ago, before digital photography became common, I ran tests with professional juries who were asked to determine whether a result in print matched the original film for color. I also used some ancient hardware to see how a machine felt about the same question. Man and machine tended to agree, except in flesh-tones, where the humans consistently rated golden skin as being closer to the original than the machine believed, and in trees, where the machine persisted in believing that the original art's greens were dull, in spite of the popular vote that a more vivid green matched it better.
Our preference for suntanned skin is most pronounced in images of light-skinned Caucasians, who can roughly be classified as anyone with a natural hair color lighter than dark brown. Hence, we do the rainforest correction in reverse: the basic recipe, but a steeper B than A curve.
In Figure 3.11A, the lightest point known to be white is at the top of the woman's sweater. it averages 94L(1)A4B. There's no obvious dark point. The dark hair below the chin reads 23L15A10B and the skin of the neck 82L9A12B. We prefer to measure skintone in areas other than the face because of the possibility of makeup that might throw our measurements off. And, in fact, the cheeks show up as pinker than the neck, with typical values of 12A10B at various values of the L.
Figure 3.11 The strong viewer preference for a tanned complexion suggests that when confronted with a pale original, left, we should use curves that emphasize the B channel, making the face more yellow, as at right.
This face is rather light, which isn't necessarily a problem. In fact, in Figure 3.11B the skintone has gotten somewhat lighter than in the original. The big issue is that the skin is pale and pink.
The L channel move lightens the shirt and the background in the interest of getting more snap in the woman's face. The A and B curves are both steepened, but the B much more, to emphasize the yellow component of the skin and hair, hopefully giving the complexion more of a golden feel.
You may notice a little extra trick here, an unadvertised departure from the recipe, caused by the annoying complication that the shirt didn't measure as neutral. Dealing with that issue will be the theme of Chapter 4, to which you may jump now if you wish. The next section of this chapter discusses why LAB corrections seem so realistic—because they work in a manner very like that of the human visual system.