Doug Kerr
Well-known member
During a recent thread on the ColorRight MAX (a white balance color correction tool that also lets us do a "shifted" color color correction in pursuit of more beauteous skin rendition than given by that soul-less theoretical color correction), there were interesting discussions on a matter of which I was previously unaware.
As I grasp it, experience has shown that, notwithstanding the gigantic range of different skin reflective colors across the human species, almost any skin color that looks not "icky" will exhibit certain relationships (even certain ranges of numerical ratios) among its color coordinates, expressed in the CMY system. And if not, we can shift the chromaticities of the skin areas (or, for convenience, of the whole image) in pursuit of these magic values, and thus of beauteous skin.
The question then arose as to why this always seems to be expressed in the CMY coordinate system.
One answer, which is very likely true, and yet certainly not the whole story, is that this wisdom was largely amassed by those working in the demanding arena of print advertising. There, images were finally audited and manipulated on the basis of a CMY (or CMYK) model, specific to the printing process involved.
Thus, whether we are working with an image of Nicole Kidman or Idi Amin, we can lay our pointer on the forehead, refer to the same set of guidelines, and say, "Of course - that's why it looks so icky" - way too little cyan, and a bit too much yellow.
But, went the argument, for most of us, we have no functional need to work in a CMY space, and surely these "litmus tests" can be applied in an RGB context.
No, went the rebuttal, they won't work there.
But, continued the debate, (if we either ignore or explicitly accommodate the matter of different RGB and CMY spaces), any color that has a CMY representation has an RGB representation, and if the guidelines allow us to test for "ickiness" in one domain, they should in the other as well.
Well, here's the story.
Assuming we accept the practical value of the guidelines as expressed for use in the CMY context, the same underlying guidelines will of course be applicable to a color expressed in RGB coordinates. But, because of the way the math works, the guidelines aren't expressed the same, and in fact their expression is no longer "handy".
Let's for a moment consider the "default" RGB-CMY transform, the one we use if we are not intersted in exactly what RGB space and exactly what CMY space is involved. I will use a 0-1 scale for all coordinates.
That transform is:
C=1-R
M=1-G
Y=1-B
Now suppose that a tenet of our non-icky skin color test is, "C should be less than 25% of M". (That's not necessary a real tenet of the test, just one to illustrate the principle.) We can write that as:
C<0.25M
Now, will that rule work in the RGB system? There, would it be, "R should be less than 25% of G"?
Well, a quick numeric check will show that this doesn't work. Duh - of course - the RGB values work in the opposite direction.
So perhaps the rule is, "R should be greater than 4 times G"?
No, that doesn't seem to work either.
Well let's use a little algebra. (I know, the gods of art will strike us down.)
We start with the original guideline:
C<0.25M
Then we substitute from the definition of the transform, and get:
1-R < 0.25 (1-G)
After some manipulation, we get:
R> 0.75 + 0.25 G
So the rule must be, "R must be greater than 0.75 plus 25% of G".
Not nearly so handy to check out in your head.
And thus, we continue to hear of the "skin beauty litmus test" expressed in terms of the CMY color coordinate system. Not a conceptual matter at all - just a matter of convenience of use.
As I grasp it, experience has shown that, notwithstanding the gigantic range of different skin reflective colors across the human species, almost any skin color that looks not "icky" will exhibit certain relationships (even certain ranges of numerical ratios) among its color coordinates, expressed in the CMY system. And if not, we can shift the chromaticities of the skin areas (or, for convenience, of the whole image) in pursuit of these magic values, and thus of beauteous skin.
The question then arose as to why this always seems to be expressed in the CMY coordinate system.
One answer, which is very likely true, and yet certainly not the whole story, is that this wisdom was largely amassed by those working in the demanding arena of print advertising. There, images were finally audited and manipulated on the basis of a CMY (or CMYK) model, specific to the printing process involved.
Thus, whether we are working with an image of Nicole Kidman or Idi Amin, we can lay our pointer on the forehead, refer to the same set of guidelines, and say, "Of course - that's why it looks so icky" - way too little cyan, and a bit too much yellow.
But, went the argument, for most of us, we have no functional need to work in a CMY space, and surely these "litmus tests" can be applied in an RGB context.
No, went the rebuttal, they won't work there.
But, continued the debate, (if we either ignore or explicitly accommodate the matter of different RGB and CMY spaces), any color that has a CMY representation has an RGB representation, and if the guidelines allow us to test for "ickiness" in one domain, they should in the other as well.
Well, here's the story.
Assuming we accept the practical value of the guidelines as expressed for use in the CMY context, the same underlying guidelines will of course be applicable to a color expressed in RGB coordinates. But, because of the way the math works, the guidelines aren't expressed the same, and in fact their expression is no longer "handy".
Let's for a moment consider the "default" RGB-CMY transform, the one we use if we are not intersted in exactly what RGB space and exactly what CMY space is involved. I will use a 0-1 scale for all coordinates.
That transform is:
C=1-R
M=1-G
Y=1-B
Now suppose that a tenet of our non-icky skin color test is, "C should be less than 25% of M". (That's not necessary a real tenet of the test, just one to illustrate the principle.) We can write that as:
C<0.25M
Now, will that rule work in the RGB system? There, would it be, "R should be less than 25% of G"?
Well, a quick numeric check will show that this doesn't work. Duh - of course - the RGB values work in the opposite direction.
So perhaps the rule is, "R should be greater than 4 times G"?
No, that doesn't seem to work either.
Well let's use a little algebra. (I know, the gods of art will strike us down.)
We start with the original guideline:
C<0.25M
Then we substitute from the definition of the transform, and get:
1-R < 0.25 (1-G)
After some manipulation, we get:
R> 0.75 + 0.25 G
So the rule must be, "R must be greater than 0.75 plus 25% of G".
Not nearly so handy to check out in your head.
And thus, we continue to hear of the "skin beauty litmus test" expressed in terms of the CMY color coordinate system. Not a conceptual matter at all - just a matter of convenience of use.