Pixiq

The received wisdom on color balance is that, ultimately, the light that falls on the scene is what counts. Measure that perfectly, and have a workflow that is perfectly calibrated, all the way through your monitor and to the printer, and you’ve cracked it.

Well, yes and no. I sense some hackles rising. What could possibly be wrong with measuring the color of light properly?  And didn’t I write an article about doing exactly that with a Gretag Macbeth card to calibrate the camera? Funnily enough, there isn’t really a conflict between making accurate measurements and sometimes disregarding them. It’s just that there’s a final piece of equipment that doesn’t respond too well to calibration — the primary visual cortex. So, you can have accurate color by using the appropriate white balance when shooting, or finding it during raw conversion, but it won’t necessarily look perfectly right. If you’re copying a painting or shooting a commercial packshot, there’s no doubt that accurate will do just fine, but most photography is far more subjective — and so it should be.

This is about tempering the natural desire to get something correct by the numbers, with having it look as you think it should. Looking as it should is another way of saying ‘according to memory’. This can mean your memory of how a scene looked at the time, or your memory of how things like that look in general. The only reason why this sounds vaguer than degrees kelvin is that we don’t have a ready and on-the-spot way of measuring our own visual response to color. There’s no iPhone app for it yet (and I hope never will be). Nevertheless, so-called memory colors are valuable tools for processing images.

Certain kinds of color are more fixed in our minds than others, meaning that we have finer discrimination and stronger opinions about whether they are accurate or not. To an extent this varies from person to person, and depends on particular interests, but there are a few standard sets that most of us recognize. They are known as memory colors because they seem to be embedded in our visual memory, and when we’re dealing with them we can usually easily say whether they are ‘right’ or ‘wrong’, and to what degree, and we can do this intuitively. Measurement is important in optimisation, but as color is ultimately a matter of perception, never underestimate the importance of subjective judgment. If it looks right, it is right.

The most valuable memory color is plain old gray, and many, though by no means all scenes have a possible gray somewhere. The great value of using gray when processing is that it is clickable — with the gray dropper. For the average run of scenes, the following are potential grey points: concrete, steel, aluminum, automobile tyres, asphalt, thick clouds, fog and mist, shadows on white (as in the folds of a white shirt, or on a white wall, but there is also a risk of these picking up colored reflections). If your shot contains one of these, good. Not all scenes have a neutral point by any means (a typical sunset, for example), but these tend not to be color-critical in any case. And, of course, you don’t want to over-correct scenes in which the color of the lighting is attractive or contributes in some way, such as a landscape lit by a low sun, or a domestic interior at night that you want to look warm and inviting.

The next most important memory color is skin tone. Complex though the colors are, almost everyone can spot even a slightly wrong color shift, and do it innately. And yet there is a seeming paradox in that the full range of human flesh tones is enormous. Not only that, but it divides into a few quite different core groups — a question of anthropology, really. So, we best know the flesh tones that we see every day.

Dedicated memory-color software such as iCorrect Portrait from PictoColor tackles this huge variety by using a relatively broad special color space, and by allowing the user to sample several colors at once. The iCorrect approach is a little unusual in that it uses a special third parameter to define color after the expected hue and saturation. This is ∆Hue, a measure of how much the hue angle varies with brightness — essential with skin.

This is a typical Caucasian skin tone range of the kind used by iCorrect:-

Other memory colors are green vegetation and sky tones. It also depends on the environment in which you grow up – if you live in New Mexico or North Africa, you will be more attuned to the colors of sand and rock than if you live in northwest Europe. Here’s a simple demonstration of our sensitivity to the color of foliage — in this case, backlit grass in a Mississippi plantation mansion. The hue difference between the main picture and the swatch is only 5º, but there’s an instant impression of rightness/wrongness. Grass, our memory tells us, shouldn’t look even slightly blue.