How to adjust skin tones in RGB

Tanuki

I really enjoyed this tutorial on Smugmug for adjusting skin tones in CMY color space. As a PSE4 user, I don't have ready access to CMY values as part of my workflow. I'm not ready to purchase the full version of PS yet (I'd rather spend the money on a lens, flash or other toy), so I'm wondering if there was a rule of thumb to accomplish this in RGB, even if the method is not quite as good as in CMYK.

On a related subject, Adobe Lightroom Beta 4.1 also does not give CMYK values. Several people have posted in the Adobe Lab Forums requesting support for CMYK at various levels (i.e., from simply providing color picker values all the way to full blown CMYK color space mode). Generally the response has been negative to all levels of requests. I personally find it ridiculous that Beta 4.1 has CMY sliders for Hue, Saturation and Luminance, yet they can't tell us the CMYK values on a pixel. It's enough to make me download a trial version of Aperature. Take that, Adobe! :flip (I didn't mean it, Adobe. I love your products, and someday soon I'll probably upgrade to CS3. :thumb)

Where was I? Oh, yes... RGB color space. Any suggestions for us PSE users?

Regards,
Mike

Duffy Pratt

If you are doing it by the numbers, there has to be a different answer for each RGB color space. The same numbers in Adobe RGB will yield more vivid colors than if they were rendered in sRGB, and both would literally pale in comparison to ProPhotoRGB or WideGamut RGB.

I did a google search for "RGB CMYK Conversion Calculator" and came up with a number of calculators, but none of them took into account either the definition of RGB or the way that the separation is done into CMYK. (CMYKs vary based on how much black ink is allowed, what the dot gain is, and the quality of the ink to be used.)

Duffy

Tanuki

Duffy Pratt wrote:

If you are doing it by the numbers, there has to be a different answer for each RGB color space. The same numbers in Adobe RGB will yield more vivid colors than if they were rendered in sRGB, and both would literally pale in comparison to ProPhotoRGB or WideGamut RGB.

We're working with Smugmug, so it's gotta be sRGB.

I did a google search for "RGB CMYK Conversion Calculator" and came up with a number of calculators, but none of them took into account either the definition of RGB or the way that the separation is done into CMYK. (CMYKs vary based on how much black ink is allowed, what the dot gain is, and the quality of the ink to be used.)

Duffy

Based on my read of the Smugmug tutorial, it doesn't really matter how much black ink is allowed, because it's based on ratios of the CMY values. The K value doesn't even factor into it. In the tutorial's example, the K value is actually 0.

I do appreciate your suggestion of using a conversion calculator. I imagine it would be a challenge to control yellow and magenta values with RGB sliders, but I'm sure it can be done.

Still I'm sure that Smugmug has been asked this question before by PSE users like me who aren't able to apply the tutorial. Surely someone has already thought this through before...

Mike

Duffy Pratt

Tanuki wrote:

I imagine it would be a challenge to control yellow and magenta values with RGB sliders, but I'm sure it can be done.Mike

It's not as big a challenge as you might think. Remember that Magenta and Green are opposites. So are Yellow and Blue. Adding blue kills yellow, subtracting blue boosts yellow. Same with green and magenta (and red and cyan). If the K is not an issue, then RGB is basically the same as CMY.

Duffy

DavidTO

Duffy Pratt wrote:

It's not as big a challenge as you might think. Remember that Magenta and Green are opposites. So are Yellow and Blue. Adding blue kills yellow, subtracting blue boosts yellow. Same with green and magenta (and red and cyan). If the K is not an issue, then RGB is basically the same as CMY.

Duffy

Keep going....I want to see you come up with a formula for skin tone ala the well-known CMYK values for skin.

Tanuki

DavidTO wrote:

Keep going....I want to see you come up with a formula for skin tone ala the well-known CMYK values for skin.

Me too. I don't think it's as easy as driving a car in reverse. But if it is, I'll be happy to be proved wrong!

One of these days, I need to read up on the theory of color spaces, ICC profiles, etc. Anyone know any good online references or books for someone with a strong mathematical background?

edgework

Interesting challenge, particularly for someone like myself who learned color in the days when CMYK ruled, and who still adjusts RGB skin tones while watching a CMYK readout.

Here's the beginning of a decidedly unscientific survey with four samples in both RGB and CMYK. For the record, the CMYK profile is Photoshop SWOP v2 and the RGB is sRGB. True, were I to convert back and forth from one profile to a different profile, I'd end up with different readouts, but, as is the case with thinking in CMYK, the proportions between plates are what is important, not the specific numbers.

Assume the fleshtones are tanned caucasian in which yellow is considerably higher than Magenta, but not out of range of what Dan Margulis says is typical.

Highlight:
5C 22M 34Y
242R 204G 168B

Quartertones:
C9 M36 Y44
235R 175G 142B

3/4 tones:
20C 55M 73Y 4K
201R 128G 83B

Shadow:
33C 83M 100Y 39K
124R 52G 20B

On the CMYK side, the distance between plates decreases as the tones darken. In the highlight we have C at about 23% of Y half again as strong as magenta.

Quartertone has C at 20% of M and Y about 20% higher than M

3/4 tones have C at 36% of M and Y about 30% higher. Black has started to creep in as well.

The shadow (dark) gives the strongest portion of C, which is to be expected as we want shadows to be more neutral. It reads 40% of M while Y is about 20% higher. The big news here is the heavy concentration of black, of course.

In the corresponding RGB values, running from the lowest number to the highest (darkest to lightest) there each plate is about 20% - 30% darker in succession.

In the 3/4 tones the difference is around 50% darker, and in the shadows, the difference is around 2 1/2 times as dark from one to the next, which makes sense since the three channels have to account for the weight that the Black plate handles in CMYK.

Rather than looking for RGB values to memorize, running through comparisons of plate values taken from "good" fleshtones will build up an instinct for the kinds of relationships that will work. For me, it's going to take some time: I look at those numbers, I understand what the percentages are saying but it doesn't mean a lot, because I'm still translating from a different language. Fluency comes with practice.

Tanuki

edgework wrote:

Here's the beginning of a decidedly unscientific survey with four samples in both RGB and CMYK. ...

Thanks, Edgework. I'll play around with this and see if it helps.

Tanuki

Okay, so I've been doing some long overdue reading about color spaces, and I can't believe what I've been missing. I've decided that it's time for me to upgrade to PS CS3 when it comes out, since I have clearly outgrown PSE.

But back to my original question about how to apply this tutorial in RGB space, I think I have a simple answer. Now, this answer is predicated on the assumption of a color space that doesn't use any black (it's really just CMY), and the CMY space is identical to the RGB space that you're working in. This allows a simple mapping of C=1-R, M=1-G and Y=1-B. I don't think this is a bad appoximation, since the example in the color tutorial shows K=0 on the color picker, and the mapping I mentioned above is accurate to within a few percent.

Okay, so here goes my attempt at converting the tutorial rules of thumb into RGB. This is just preliminary, since I still need to check it against the sample photos on the tutorial.

90% of all you need to know is that you can never let the yellow % fall below magenta % on anyone's skin unless you're trying to show sunburn.

"... never let the blue value rise above the green value".

A fair-skinned pinkish baby could be as light as 15% magenta, 16% yellow.

"... could be as light as 218 Green, 214 Blue".

Most caucasians fall in the range of 5-20% more yellow than magenta.

Is this talking about absolute percent? If so, then this equates to "in the range of 13 to 51 levels less blue than green."

A fair-skinned caucasian adult could be as low as 20% magenta, 25% yellow.

"... as high as 204 green, 191 blue."

A bronzed caucasian could be as high as 45% magenta, 62% yellow.

"... as low as 140 green, 96 blue."

Asians and hispanics have typically 10-20% higher yellow than magenta.

Once again assuming that this means absolyte percent, this equates to "... 26 to 51 levels lower blue than green."

If magenta is getting towards 50%, beware of the well-done look.

"If green is getting down towards 127, beware of the well-done look."

On pleasing photos, cyan usually falls between 30% to 50% of the magenta value. Less than 30% of magenta makes sunburn; more than 50% of magenta makes makes them ghostly blue.

Edit: I have to work on this one. Stay tuned...

Any comments are welcome.

Mike

imann08

I am not all that familiar with PSE but assuming that it has an identical info pallete, wouldn't it just be a matter of making the upper left numbers set to actual value and the upper right values set to CMYK. That's how I do it at least. It doesn't matter what RGB profile you are using.

Of course, it helps to know what was mentioned before, that the blue is opposite of yellow etc etc. The more you do it the more you will grow accustomed to what the right values should be and then you can verify them with the CMYK values in the info palette.

Tanuki

imann08 wrote:

I am not all that familiar with PSE but assuming that it has an identical info pallete, wouldn't it just be a matter of making the upper left numbers set to actual value and the upper right values set to CMYK. That's how I do it at least. It doesn't matter what RGB profile you are using.

Unfortunately PSE does not have CMYK... not even in the info pallete.

rutt

Probably no LAB either. That wold also work.

Tanuki

rutt wrote:

Probably no LAB either. That wold also work.

Elements has RGB and that's it.

This will be a moot point for me once I get CS3, but thought it would be nice to work out the RGB rules of thumb for the Elements users that I'll be leaving behind.

jfriend

Tanuki wrote:

Elements has RGB and that's it.

This will be a moot point for me once I get CS3, but thought it would be nice to work out the RGB rules of thumb for the Elements users that I'll be leaving behind.

Also useful in ACR or Lightroom that also won't show CMYK numbers.

Tanuki

jfriend wrote:

Also useful in ACR or Lightroom that also won't show CMYK numbers.

Oh yeah, I forgot about that. The fact that Lightroom doesn't have CMYK values is what actually got me started on this little project. I had posted to the Adobe Labs forums to request the feature, but the idea was poo poo'ed pretty strongly by other participants in the forum saying that full CMYK support won't happen in version 1. I tried to explain the difference between full CMYK support and providing approximate CMY values in the info tool, but it fell upon deaf ears. I mean, shouldn't one expect Lightroom to give useful control over skin tone without having to open PS? Oh well, maybe these RGB rules of thumb can be worked out and it will become a moot point.

DavidTO

Tanuki wrote:

I mean, shouldn't one expect Lightroom to give useful control over skin tone without having to open PS?

Yes, this was one reason I was using Capture One for so long. CMYK eyedropper.

jfriend

Tanuki wrote:

Oh yeah, I forgot about that. The fact that Lightroom doesn't have CMYK values is what actually got me started on this little project. I had posted to the Adobe Labs forums to request the feature, but the idea was poo poo'ed pretty strongly by other participants in the forum saying that full CMYK support won't happen in version 1. I tried to explain the difference between full CMYK support and providing approximate CMY values in the info tool, but it fell upon deaf ears. I mean, shouldn't one expect Lightroom to give useful control over skin tone without having to open PS? Oh well, maybe these RGB rules of thumb can be worked out and it will become a moot point.

Yeah, I'm pretty disappointed that they didn't understand the request. I've made the same request for ACR before. I someitmes have to tweak WB, open in PS, look at the CMY numbers, close without saving, tweak in ACR again, etc... just to try to get the skin tone right in ACR. For this reason, I'm following the RGB discussion and I'll have to see if I can write it down and then start trying to use it.

BinaryFx

Does Elements have a "Color" window/floating palette or tab or other such area? Being RGB, the numbers rely on the editing space in use (which may not be a concern for PSE users, but for others sRGB requires different numbers than Adobe RGB for the same colour).

If one samples an image colour, the foreground colour swatch in the toolbox will change. Now for the key point. If there is also a Color palette or tab, with R, G and B sliders that allows one to mix colour then there is a nice trick...does Elements have such a feature? If not this is handy for full version users. For CMYK work, it can also be good to set the lower display area to grayscale ramp, rather than RGB or CMYK spectrum. One can then mouse over and click and drag along the gray gradient and the CMYK sliders will reflect the gray balance for the particular profile embedded/assigned/presumed for the image at hand.

Back to skin tones. There is a visual relationship between the position of the sliders - which reflects the numerical ratio of C to M to Y or R to G to B that are being discussed as an idealised aimpoint (see the image below). One can also change the Color palette to mix values in CMYK or LAB colours in the full version which can also be handy for visual evaluation using this slider ratio position (LAB sliders are really good for this).


Regards,

Stephen Marsh.
http://members.ozemail.com.au/~binaryfx/
.

Tanuki

BinaryFx wrote:

Does Elements have a "Color" window/floating palette?

If one samples an image colour, the foreground colour swatch in the toolbox will change. Now for the key point. If there is also a Color palette or tab, with R, B and B sliders that allows one to mix colour then there is a solution...does Elements have such a feature?

Thanks for the suggestion. I can't find this feature in Elements, so I'm assuming that it doesn't have it. If anyone knows otherwise, please clue me in.