How to nail CMYK skin tone values BEFORE printing?

urbanaries · June 1, 2007

I have read the skin tone tute several times, and I understand the concepts, and the rules of thumb, but I don't read between the lines well when there are numbers involved. I also hear the praises of auto color, and have had good results from it if I do *nothing* to the images out of camera. But when I make saturation, contrast, white balance or exposure adjustments in postprocessing, which I often do, sometimes auto color on top of that yields some funny color shifts. So instead of the guessing game, I think I need to perfect my skin tone values in post, and use True color. (please correct me if I'm wrong).

Basically, I don't want to beg for Andy's mercy after the ink has dried anymore :thumb

Ok, so here is the "edited" image, before my skintone work:

A sample from the forehead yields
C 16
M 20
Y 11
K 0

Yellow is lower than Magenta. Strike one. Her hair looks redder here than in person, too.

Tute says:
"A fair-skinned pinkish baby could be as light as 15% magenta, 16% yellow. Most caucasians fall in the range of 5-20% more yellow than magenta."

She's fair but not a pinkish baby anymore. So, I do I accept the CURRENT magenta value and ADD yellow? I'll say 10% more yellow than magenta for this case, as a starting point. So that puts us at M 20, Y 22.

Looks a little yellow, but i'll keep going. Maybe adjusting the cyan next will even it out.

Tute says: "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."

Ok, so split the difference and go 40%. Magenta is 20, so that puts the Cyan at 8.

Hmph. Two questions:
1. Am I interpreting and executing the tutorial correctly?
2. If so, am I color blind.....because the 3rd image looks awful. Way too greenish yellow compared to first, on my (at least I believe it to be) calibrated monitor.

Thanks for helping me trouble shoot my apparent shortcomings.

ETA: Here is fourth version, with skin fix applied to entire image, and large brush erasing the curves layer over the grass.

MarkR · June 1, 2007

I think the skin tones on the third pic are spot on. Unfortunately, your grass highlights all turned yellow, which might be what's giving you trouble with the image. But yes, for the first two pictures, the girl's skin looks too blue/gray.

urbanaries · June 1, 2007

The Hobbyist wrote:

I think the skin tones on the third pic are spot on. Unfortunately, your grass highlights all turned yellow, which might be what's giving you trouble with the image. But yes, for the first two pictures, the girl's skin looks too blue/gray.

Thanks for your response! I think you're right, the grass is throwing off my overall view of the image. And her yellow shirt isn't helping matters.

Should I tweak my values a bit? Or go back and do a layer mask of her skin and adjust that independently?

Andy · June 1, 2007

urbanaries wrote:

Thanks for your response! I think you're right, the grass is throwing off my overall view of the image. And her yellow shirt isn't helping matters.

that part is okay, the yellow reflection under-chin is sorta neat

Should I tweak my values a bit? Or go back and do a layer mask of her skin and adjust that independently?

mask is one way, and it's quite easy to do, I just use a big fat brush here, and brush away the change on the grass, leaving the baby perfect.

urbanaries · June 1, 2007

Andy wrote:

that part is okay, the yellow reflection under-chin is sorta neat mask is one way, and it's quite easy to do, I just use a big fat brush here, and brush away the change on the grass, leaving the baby perfect.

That's a great suggestion! Brushing away the grass instead of quick masking the skin would be much easier.

thanks for walking my small brain through this...

MarkR · June 1, 2007

urbanaries wrote:

Thanks for your response! I think you're right, the grass is throwing off my overall view of the image. And her yellow shirt isn't helping matters.

Should I tweak my values a bit? Or go back and do a layer mask of her skin and adjust that independently?

You could do that. Sometimes it looks fake and sometimes not.

I'm at work right now so I can't analyze the values, but you might be able to do a curve of the yellow channel that holds the highlights in place and corrects the skin values.

Love the eyes on that baby, btw.

urbanaries · June 1, 2007

The Hobbyist wrote:

You could do that. Sometimes it looks fake and sometimes not.

I'm at work right now so I can't analyze the values, but you might be able to do a curve of the yellow channel that holds the highlights in place and corrects the skin values.

Love the eyes on that baby, btw.

thanks, it was hard to get her to open her eyes WIDE!

I appreciate your suggestion but have no idea how to hold the highlights in place with a curve.

As you can tell this isn't exactly my forte.

Updated a fourth version in OP, that way its easier to compare visually...thoughts?

And the Big Question Andy....TRUE or AUTO?

Andy · June 1, 2007

urbanaries wrote:

And the Big Question Andy....TRUE or AUTO?

I don't say it much, but True for this one.

urbanaries · June 1, 2007

Andy wrote:

I don't say it much, but True for this one.

Is that a good thing or a bad thing?

Is it that I've screwed with the image so much, that auto wouldn't evaluate it properly?

Or is it really that I get a gold star for the colors being accurate enough?

Seriously. I am trying to learn how to fish here so you don't have to bait my hook for me everytime....

(you've told me True before, but I was never sure if it was because the colors were beyond help.)

arodney · June 1, 2007

CMYK is a print, output color space. It's device dependant meaning that every CMYK printer will require a different mix of CMY and K values to produce the same color appearance. What CMYK device are you printing to? If you have an ICC profile that fingerprints this device, it will produce the correct values from an RGB document.

There's this old and pretty silly myth that there's a general recipe for CMYK for skin (or other memory colors) that people should adopt. It dates to the old days when highly skill drum scan operators would target values for CMYK to THEIR devices. IOW, over the years, they knew that a fixed mix of CMYK values would produce acceptable color appearance for their specific printers (usually a press of course). Then these people would tell others to use the values to try to fix things by the numbers. They failed to tell people not using the same device this wouldn’t work. And as I pointed out, getting values today is easy if you have a good device profile.

Better is to simply look at the color on a calibrated and profiled display!

One other trick is to output an image of known quality to the device (a print you have achieved good skin tones). Then open that image up as you correct others for a visual reference.

Lastly, look at all the differing skin tones around us (Caucasian alone) and then consider how rather silly it is to attempt to place fixed numeric values on this, and doing so based no an output device (CMYK) is a huge, huge stretch.

urbanaries · June 1, 2007

Rodney, the color/skin tone issue seems so simple when others explain it, yet as you've pointed out, there are so many variables.

Calibration is not foolproof. Nor are my eyes. I have no idea how you interpret color versus me. How else do the color blind realize they are color blind? I suppose the ICC profile plus calibration, plus CMYK values are the best I can do, but to be honest it still doesn't give me assurance. I still feel like I'm driving a car with a blindfold on.

I understand now why several portrait photographers claim they "specialize in black and whites."

I also see why photographers invest in printers themselves, so they can monitor these variables quickly and consistently.

Its enough to pull my hair out, literally.

ETA: I don't think smugmug is providing fixed values for skin tones, but rather ratios.....

DavidTO · June 1, 2007

Great shot to play with, but the best solution, IMO, for work like this is to set a custom white balance, or to set it in RAW conversion. I think there's something that went wrong before image #1.

urbanaries · June 1, 2007

DavidTO wrote:

Great shot to play with, but the best solution, IMO, for work like this is to set a custom white balance, or to set it in RAW conversion. I think there's something that went wrong before image #1.

thanks for your input David....I did set the WB in RAW conversion, but apparently not correctly.

DavidTO · June 1, 2007

urbanaries wrote:

thanks for your input David....I did set the WB in RAW conversion, but apparently not correctly.

How did you set it? Maybe it was right. It just seems to me that if it's right, the skin tones shouldn't be so difficult to get right.

EDIT: try 5000k and tint -5. That's Marc Muench Daylight.

pathfinder · June 1, 2007

Lynne,

Lots of green reflectors in this image - green grass and a green interior of the tub.

Are you sure there is no extra green light on her face in the original shot? I'd be willing to bet there is a lot of extra green on her face, at least prior to processing. Extra Green will kill any magenta present...

Neutral props are not as attractive as pink and green ones, but they do not interfere with color balance issues either.:D:D

urbanaries · June 1, 2007

DavidTO wrote:

How did you set it? Maybe it was right. It just seems to me that if it's right, the skin tones shouldn't be so difficult to get right.

EDIT: try 5000k and tint -5. That's Marc Muench Daylight.

OK, here's SOC with MMD. I typically shoot in AWB unless I'm gelling my flash indoors, and then adjust in LR by eye (obviously not getting great results w/ that method.) But I'm not sure MMD works here, in the shade.

urbanaries · June 1, 2007

pathfinder wrote:

Lynne,

Lots of green reflectors in this image - green grass and a green interior of the tub.

Are you sure there is no extra green light on her face in the original shot? I'd be willing to bet there is a lot of extra green on her face, at least prior to processing. Extra Green will kill any magenta present...

Neutral props are not as attractive as pink and green ones, but they do not interfere with color balance issues either.:D:D

DUH!!! thanks PF for conneting those dots....see original and yes there is tons of green, i figured it was from the grass, didn't even consider the prop.

DavidTO · June 1, 2007

Also might try Shade WB?

urbanaries · June 1, 2007

DavidTO wrote:

Also might try Shade WB?

that looked really yellow, but might hve been a better starting point.

DavidTO · June 1, 2007

Also mebbe use MMD as a starting point, but adjust the tint to offset the green? Dunno. I'm at work and unable to substantiate my thoughts in PS.

pathfinder · June 1, 2007

Is the RAW file available online somewhere??

pathfinder · June 2, 2007

So -- Lynne, Sorry I was at work with an uncalibrated lousy LCD and I do not trust image evaluation on it. This post was composed tonight on my Apple LCD that is a calibrated device.

I worked from your first image posted here in its smaller size here on dgrin, since the larger files seemed unavailable.

I loaded it into PSCS3 and ACR4.1 as a jpg. I used the temp AS SHOT in the Adobe Raw Convertor (which will open jpgs in ACR in PSCS 3)

Looked like not enough yellow as you said and maybe too much cyan as well, so I used two curves as shown below...

Pulling down the Blue channel slightly will increase the yellow in the image as needed

But the image also had slightly to much cyan, which will be reduced by adding red as shown below

The result of these two curves is the following image which has pretty good CMY numbers on her cheeks, forhead and the top of her left hand

I did not mask out the grass which you may now feel is too yellow, but the curves I have posted should help you get better numbers for her skin tones, and they look reasonable on my monitor.

Is this helpful at all??

urbanaries · June 2, 2007

pathfinder wrote:

So -- Lynne, Sorry I was at work with an uncalibrated lousy LCD and I do not trust image evaluation on it. This post was composed tonight on my Apple LCD that is a calibrated device.

I worked from your first image posted here in its smaller size here on dgrin, since the larger files seemed unavailable.

I loaded it into PSCS3 and ACR4.1 as a jpg. I used the temp AS SHOT in the Adobe Raw Convertor (which will open jpgs in ACR in PSCS 3)

Looked like not enough yellow as you said and maybe too much cyan as well, so I used two curves as shown below...

Pulling down the Blue channel slightly will increase the yellow in the image as needed

But the image also had slightly to much cyan, which will be reduced by adding red as shown below

The result of these two curves is the following image which has pretty good CMY numbers on her cheeks, forhead and the top of her left hand

I did not mask out the grass which you may now feel is too yellow, but the curves I have posted should help you get better numbers for her skin tones, and they look reasonable on my monitor.

Is this helpful at all??

Thanks for this detailed explanation! I will read it a few more times, but think I get what you did, and it does help me understand the relationship between the three color values better. The one question I do have is where did you get the points on the curve, and am I correct in seeing those points are different locations in each curve?

thanks again and feel free to PM answers so not to dumb down your forum to infancy.

pathfinder · June 2, 2007

Lynn,

I did not set the points by looking at the numbers so much as run my mouse pointer over the child's face and identify where the points on the child's face resides on the blue and red curves. I then pulled down the blue curve a bit ( raised yellow in other words) and raised the red a bit ( killed some cyan ) by eyeballing the image on my monitor. I then read several different points on the forhead, cheeks, and hands until I was satisfied that they were within line for Caucasian skin. If they were not right, I could either increase or decrease the changes I made in the curves.

This is not the only way to approach this problem, but it was simple and straightforward to do and explain.

Your numbers do not have to match mine exactly, but just look reasonable on your monitor and when examining the RGB numbers in the skin tone areas.

You could of course mask out the changes in the grass if you prefer it unchanged.

MarkR · June 3, 2007

This might be a case for doing some of Dan Margulis's channel-blending techniques from his Professional Photoshop book.

I tried my hand at blending the green into the blue in lighten mode to try to kill the strong yellow grass. Here's my attempt:

Comments and criticism welcome.

[edit:] did some quick curves to increase the reds and yellows in the face.

jjbong · June 3, 2007

The Hobbyist wrote:

Comments and criticism welcome.

I prefer pathfinder's, as I find the contrast in the face and the clothing
attractive, and you've lost this.

Is the yellowish grass in the background really distracting? It might be,
I'm not sure.

Anyway, here's quick mod to pathfinder's using selective color to reduce
the yellows in the grass. I did a selective color on yellows (which is what Photoshop considers the grasses to be), absolute, Y -13. It tones down the yellows in the grass a bit, without toning down the cloths too much, I think.

MarkR · June 4, 2007

jjbong wrote:

I prefer pathfinder's, as I find the contrast in the face and the clothing
attractive, and you've lost this.

Fair enough. A quick move to lab to increase contrast in the L, and a quick saturation boost gives me this:

I am intrigued by the selective color tool, but haven't really used it/understood it fully. Wouldn't reducing the yellow in yellow also affect the yellows in the girl's skin/clothes?

jjbong · June 4, 2007

The Hobbyist wrote:

I am intrigued by the selective color tool, but haven't really used it/understood it fully. Wouldn't reducing the yellow in yellow also affect the yellows in the girl's skin/clothes?

Your latest version looks pretty good. I also played around with blending your grass in with pathfinder's baby (see end of post). I did this with a blend-if in LAB, setting the white sliders on the A channel to 119/124.

Regarding selective color, I've seen Dan use it for quick fix-ups at the end of working on a picture (to get cyan out of the grass or sea, for example). It will hit the child's dress (the skin to a much lesser extent), which is why I couldn't
push the effect to the extent that you were able to do. You can get a pretty quick read on what parts of the picture selective color thinks is yellow (or whatever you've selected) by pushing the Black slider all the way to the right. The areas affected for your color selection will all darken noticeable.