A Closer Look
Politically insensitive jokes about vision deficiencies are exceedingly common in the publishing industry—normally with respect to visual (if not mental) impairments allegedly suffered by clients who offer certain questionable opinions about color.
Even the best of us can suffer this indignity. Once when I supervised a pressrun, the pressmen, offended by some of my decisions the previous time around, announced that the only reason I was being permitted to do it again was that Stevie Wonder had turned the job down and Ray Charles was out of town.
Calling other people blind is a frustrated reaction to what everybody knows: we all see color differently and therefore we all have different, sometimes radically different, ideas of what looks good.
One of the most irritating challenges we face today is determining what constitutes a visual match. So, without further ado, you are hereby assigned to make such a decision. In Figure 3.12, #1 is the original. Which of the other four do you think is the closest match to it? Which the worst?
Figure 3.12 The question is, supposing that #1 is the reference image, which of the others is the closest—and which is the worst—visual match to it?
This question isn't as weird as you might think, provided you visualize #1 as being held in your left hand while the rest of the book is held in your right. Say #1 is found in an annual report that uses much whiter paper than this book does, for example. In that case, we wouldn't be able to match it exactly. Nevertheless, we'd try to get close—and choosing the best way to do so is not as easy a call as this one appears to be.
I empaneled a jury of 12 to come up with the answer here and in several other sets of images. You probably won't agree with their verdicts, particularly (speaking of politically insensitive commentary) if you are female.
Justice may be blind, but this jury pushed the envelope. I decided to find out how bad it would be if clients were actually as visually impaired as many retouchers and printing firms claim. To that end, I empaneled a jury of the color-blind. And color blindness is almost exclusively a male phenomenon.
The Protanope-a-Dope Strategy
Color blindness is caused by a sex-related recessive gene. Men have only one of these genes, and the wrong one turns up around eight percent of the time. Women have two. If either one is normal, the woman has no problem herself, although if the other is bad, half of her male children will be color-blind. For the same reason, men are vastly more likely to be afflicted by the disastrous blood illness hemophilia. According to my wife, sex-related recessive genes also exist for dishonesty, shallowness, poor listening skills, and inability to keep a kitchen clean, although the scientific evidence to support her position is as yet inconclusive.
The term color-blind is misleading, because those afflicted have no problem seeing many colors. Plus, some individuals have more of a deficiency than others.
The overwhelming majority of the color-blind, including all members of my jury, are said to suffer from red-green color blindness. This, too, is a misleading term. Magenta-green is a more accurate description. I showed the jury the famous Kodak picture of Figure 3.13 and asked what color the woman's hat was. All but one responded "Brilliant red."
Figure 3.13 The common term for the affliction is the inaccurate "red-green color blindness." A group of such individuals nevertheless had no difficulty saying what color this woman's hat is.
There are at least two categories (some say four) of magenta-green deficiencies. Part of my testing was a protanope-and-deuteronope-adope strategy aimed at finding which category each juror belonged to. It turned out that the class of deficiency played almost no role in the men's evaluations of the images.
If individuals with normal vision were evaluating Figure 3.12, I'd expect a unanimous vote for #3 as best match to #1, and #5 as worst match. I would also expect, but not guarantee, a decision that #2 is closer to #1 than #4 is. The color-blind jury, of course, felt very differently.
This was one of a series of similar comparisons, all produced in LAB. The black numbers were burned into each image so that they could conveniently be viewed on screen.
In each case, I created one alternative by horsing around with the L channel, while doing nothing to the AB. That version (#3 in this set) therefore matched the original for color but not for detail. If you're normally sighted, that's probably your favorite, because the color in all three alternatives is so abysmal. But we can't pretend that it actually matches.
There was always a second version (#4 here) where I flattened the A curve in just the same way that we've steepened it in every example so far. That is, the center point stayed the same, but instead of rotating the curve counterclockwise around it, I rotated it clockwise. This savagely reduced saturation along the magenta-green axis.
In a third version (#2 here) I flattened the A only half as much, but I also modestly flattened the yellow-blue B channel. And there was always a trick image, #5 here, that was meant to prove a point and about which I asked specific questions. The trick in Figure 3.12 is that #5 is identical to #4 except that the A has been inverted; that is, everything that tended toward magenta now tends toward green and vice versa.
The jury always hated the version that toned down the yellow-blue channel. If the colors of the image were relatively subdued, as in this river shot, the jurors tended to like the version that matched for detail but washed out the magenta-green distinction. In more brilliant shots, they started to abandon the version with the unchanged L in favor of the one with true colors. The only one in which the wrong-detail picture decisively beat the magenta-green washout was this Kodak image, where the hat is apparently so red that even the color-blind take notice.
Since Figure 3.12 doesn't feature many bright colors, only three jurors voted for #3 as the best. Yet only three went for the normal winner, #4. The other six rated it an unusual kind of tie.
I asked for a description of the differences between #4 and #5. Half the jury replied that the two are identical. Five people saw significant differences, but only one offered a description reasonably similar to what a normally sighted person might say. The others made comments like "There is a saturation issue in the trees."
Those who thought there was something seriously messed up about #5, even without knowing what it was, naturally voted it as the worst. The other six did the conventional thing for the color-blind, choosing #2 as worst match.
Those who saw #4 and #5 as identical all voted the two of them as best match to #1. This makes sense. These people can't see the magenta-green component at all, so they completely ignored the flat-looking colors of #4. They only saw that #4 and #5 match the detailing of #1 and that #3 doesn't.
One of the most valuable attributes anyone working in the graphic arts can develop is the ability to predict how clients may react to certain images. Given what you know now about this color-blind jury, how do you think it felt about the Halloween image of Figure 3.14? Once again, image #1 is the reference version. You are supposed to say which of the other four matches it the best, and which the worst.
Figure 3.14 Another challenge to the color-blind jury: assuming that #1 is correct, which of the other four is the closest visual match to it, and which the worst?
It Can't Be Put into Words
The variations here follow the same lines as before, with the numbers shuffled. Version #2 corresponds to #4 in Figure 3.12: the magenta-green component is sharply reduced by a severe flattening of the A. Figure 3.14's #3 is the equivalent of Figure 3.12's #3: the color is the same, but there's a slight change in the L, causing a variance in detail. The overall reduction in all color that was #2 in Figure 3.12 is #5 in Figure 3.14. #4 is the trick image, a specialized treatment of the A channel in which darker magentas, such as the man's jacket, were left alone, but softer colors, such as his face, were grayed out. And I asked for specific comments comparing #4 and #5 of Figure 3.14.
With a single exception, the jury made its normal choice for the best, splitting between #2 and #3. Version 5, with its diminution of the yellow-blue component that the color-blind see well, was, as usual, voted worst, but there were also several votes for #4 as worst, presumably from those who could see enough of the magenta component to have a serious issue with the man's face.
The group had a hard time evaluating the man's jacket in #4 and #5. Let me paraphrase some of their comments:
- One is lighter.
- One is more saturated.
- One is punchier.
- One is browner.
- There's no difference at all. As for the clown's costume, most of the jury correctly said that it was more yellow in #4. But, strangely, four jurors said that in #5, it had turned green.
Accordingly, if you're thinking of designing graphics with the color-blind in mind, you have to accept that there's a lot of variation in the group. These jury findings suggest that the best way to predict how a color-blind person would react to a picture is to convert it to LAB and apply the curve shown in Figure 3.15—the same curve that produced version #4 of the river scene and #2 of the Halloween image. Keeping that in mind, how do you think the jury voted in the simplest of all these sets, the vividly colored Figure 3.16?
Figure 3.15 The quickest way to approximate how a color-blind person might see an image is to apply this curve to the A channel.
The variant of #1 that uses this curve is #2, and that's the version that the color-blind think is a closer match than #5, which seems so obviously closest to the rest of us that we don't notice that its darkness doesn't quite match.
The jury always disliked anything that hurt the B channel, like #3 here, which is analogous to #2 in Figure 3.12 and #5 in Figure 3.14. I offered them a choice of poisons by generating #4, in which everything red has been wiped out (damaging the B channel as well as the A), while pure blues and yellows were left alone. The vote for worst match to #1 was a dead heat between these two versions.
Using Figure 3.15's curve is a good deal more accurate than just assuming that people are "red-green color-blind," or that they see a black and white world, but it has severe limitations, in that nobody has ever been both color-blind and normally sighted in the same lifetime. Thus, no eyewitnesses can be called.
The conventional wisdom is that those who are deprived of one sense compensate by becoming more acutely sensitive in the others. Do we know how a symphony of Beethoven sounds to a blind person? Or whether deaf people are able to see things in Renoir that the rest of us cannot?
There was only one unanimous response to any of the questions. In an ocean shot, the water was not particularly vividly colored. I asked in which version it was the least blue. The question was difficult enough to stump some normally sighted persons. Every member of the jury nevertheless got it right. They appear to respond to blue at least as well as the rest of us do. Is it possible that they actually see it better?
Nobody has any trouble detecting that the clown's costume is less yellow in #5 than in any of the other variants in Figure 3.14. But what do you make of the strong minority sentiment that it's gone green? I myself don't see greenness, yet by measuring the numbers that make up the jacket, one can construct the argument that it is in fact more green. Can the color-blind be picking up a distinction that's too subtle for the rest of us? We can see colors that aren't even on their map, but is it not possible that the converse may also be true—that the green they are reporting is simply a color that's out of our gamut?
In that Kodak picture of Figure 3.13, are the color-blind just seeing a duller-colored fleshtone, or do they detect more yellowness as well? Do they see her hair as a brilliant yellow, rather than the mousy eyesore the rest of us perceive?
And What If God Is Color-Blind?
In 1994, in the first edition of Professional Photoshop, I showed examples of the work of a colorblind person who had been trained to correct images, by the numbers perforce.
That sideshow received more attention from readers than it actually deserved. But it was a measure of how we are fascinated by the way other people perceive things.
Some of us are fortunate enough to create images that need please only ourselves. But most professionals have to please a client, or a reader, or an art director, or somebody else whose preferences—and perhaps, whose ability to see certain colors at all—is in question. And everyone who has been there has found out the hard way about the consequences of these perceptual differences.
Granted, having a color-blind person evaluate your work is tough. If your client tells you that #2, #4, and #2 are the best matches to #1 in the river, Halloween, and hockey images, respectively, it can be a little disconcerting. Yet this is merely an extreme case of what happens everyday in the real world. As people age, their corneas get yellower and they lose perception in certain shades. Also, many drugs, Viagra being a notorious example, affect color perception.
The questions listed just above the last subhead therefore are not just an academic matter. If people see color differently than we do, we can hardly blame them for acting on it. And who knows, perhaps they're right.
Having our own opinion about these matters is both natural and desirable, provided we don't take it too seriously. And, given the differences in human perception, we have to realize that the client is king. After all, in the land of the color-blind, the spectrophotometer isn't even the scullery maid.
At high noon in hell, when viewing conditions are the best, punishment is meted out to the unholy. Every color management consultant who has overhyped his technology or overrelied on the measurements of an artificial measuring instrument, is forced, day after day for all eternity, to argue and demonstrate the accuracy of his profiles—to a jury of the color-blind.
Perhaps it is all a matter of divine humor and spite. How do we know what is normal, and what is defective? Perhaps God Himself sees the clown's jacket as green, and has difficulty with magentas. Perhaps our own ability to distinguish these magentas and greens is a handicap, a cosmic joke, and not reality at all.
Fortunately, there's no need to decide this issue, not that it could ever be decided. We must merely concede that there is not necessarily a correct answer to the question of which version is the best match, just as there is not necessarily a correct answer as to which looks the best.
Magenta, green, yellow, blue. The wavelengths of light intrude upon us, enter our consciousness, trick us into thinking we see an absolute, laughing at us the whole while, knowing that for all the certainty we feel in our own perceptions, each of us is just a bit color-blind.