Impossible Colors and Colorspaces
What do we mean by impossible colors? Healthy foliage, of any type, is almost always slightly more yellow than green, and it is never more cyan than yellow. When foliage in an image violates this rule, we know there’s a color cast and correction work is called for (unless it’s there for artistic reasons, obviously). Similar rules of thumb exist for faces and skies.
The first tool I’m aware of that used this knowledge was the Curvemeister "Hue Clock". Depending on where the "hour hand" pointed, there were guidelines for skin, sky, and foliage. This was initially developed by comparing data in the RGB colorspace. A few years ago, Dan Margulis (a big name in color correction as you probably know), realized the Lab colorspace was actually easier to use for this.
Not that many folks are familiar with Lab. The following figure shows the relationships among RGB, CMYK, and Lab.
It’s a plot of a vs. b in the Lab colorspace. There are three planes of overlapping colors in the center for Lightness values of L=25, 50, and 75. These are included as a reference for where the colors are located in Lab space.
Note colors get Warmer as we go to the upper right and Colder as we go to the lower left. RGB and CMY are included, so it’s easy to see their relation to each other and Lab. Finally, shaded areas denote the four hue ranges where we expect certain things to reside.
Skies should have a b value of about -30 or less, and should only have a very small a value (plus or minus). Typical American-born adult skin should have a positive a value about the same as or larger than the b value. For children or folks with blue eyes, the b value must be positive and should be about the same or larger than the a value. The characterization “blue eyes” is Dan's term. Folks from Europe may also belong in the Blue-eyed section. I have some Italian friends that fall into this area. Regardless, skin from any race, age, or area should never have a negative a or negative b. People’s skins don’t have green or blue tones, period.
In practice what we do is set the second Info palette in Photoshop to display Lab. We mouse over Sky, Skin, or Foliage and see if the colors follow these rules.
- For Sky: a should be less than about -30 and b needs to be small.
- For Skin: a and b both have to positive. If a>b, typical "American born" skin is suggested. If a<b, we expect this to be a child, someone with blue eyes, or the Italians I know. We can better refine this by seeing what the values are in photos we know are properly color corrected.
- For Foliage: a must be negative, b must be positive and |a|should be less than b and greater than 1/2 b.
Comments