I'm still struggling with skin tones. At this point the dilemma is how much to trust my eye, and how much to trust the recommended numbers. I've posted three images here as an example. The first is what looks good to my eyes. The second is what has the "appropriate" CMYK numbers according to many guides (however, it has far too much of a magenta cast for my taste). The third was made using iCorrect's "1Click" plug-in. I prefer it to the second, but the CMYK numbers are off. (Sorry about the watermark - I haven't decided whether I want to buy it yet, so I'm in a trial period)

What do you think?

I guess in the end color is so subjective it only really matters what I like, unless I'm sending them in for publication (which I'm not). I do at least want to know how to get what the pros consider to be good skin tones, and then I can decide whether I like that or not!

Welcome aboard, Switters.

I think you hit the nail on the head when you said, "I guess in the end color is so subjective it only really matters what I like..." I like the 3rd one for skin tone, but her hair looks too green to me.

Yes, I found it difficult in this particular image to get the right skin tones and hair color at the same time. I could of course do a composite in Photoshop, but I'm trying to find out if there's a way to nail it in one shot. In the versions where her hair looks right her skin has too much magenta. In the version where her skin looks good, there is too much green in her hair.

Switters Personally i think skin tones are a nightmare, but i found this tutorial which i found to be very helpful. The actual "doing" bit is easy with some useful info provided
The part marked
"Is it easy to do in photoshop"
are the actual instructions
http://www.smugmug.com/help/skin-tone

Let me know what you think it would be interesting to get some one else's view on this technique


Palms

Hi,

I've seen that tutorial and found it helpful in some ways. What I've noticed is that the single most important thing in getting skin tones right, or close to right, is proper WB. When the WB is off, it's very hard to get good skin tones no matter what I do with curves, LAB, etc. When the WB is right on or close, it usually doesn't take much adjustment to get good results. So I've started using a WhiBal card when I shoot in tungsten/mixed lighting, and that has helped a lot.

Beyond that, I do my WB corrections in Lightroom. If there is a color cast I can't correct there, I export to PS CS3 and try using the eyedroppers in curves. If that doesn't work, I may then try "correcting for the numbers" as they explain in the Smug Mug tutorial. I've also recently begun correcting in LAB space, which has some advantages.

Just another thought would Cspringers digital grey card action be of any use. I don't use it, as it isn't compatible with ps7 but maybe worth a look

http://www.atncentral.com/download.htm

Palms

Hi there

Skin tone is sooo subjective. I used your second picture and by my numbers had too much magenta and cyan.

I adjusted as below... It may be just a tad yellow but the numbers I use are yellow slightly more than magenta and cyan a 1/3 of magenta. If you adjust luminosity it may change these values… I often do mostly reducing saturation a little bit.

It is very important where you are sampling color... I try to find an area neither highlight or shadow then make a selection of like areas and average the color... I make a temp layer and use it to adjust a curve.

Butch

Probably the most common problem you may come upon with people photos is too much magenta in the skin. The trick is to correct it without adversely affecting the other colors in the image.
One very effective method is the use of a Selective Color Adj layer or the Image>Adjust>Selective Color command. Select the Red channel and then just back the Magenta off a little. To prevent too much green experiment with a combination of moving the magenta down a little and the yellow up a little. Example 1 attached.

You will probably read or have red the folks who like to adjust the skin tones by the numbers. I am one of those. However, as you have noticed, the suggestions in RGB and CMYK leave things to be desired. The proportions are variable and results inconsistent and often you need to keep tweaking 3 channels.
Instead I would encourage you to try out Curves in LAB color. There is a very readily defined relationship between the a and b channels that define a range of skin colors. The L channel only affects the luminosity or brightness. Shifting the value of a and b relative to each other (both are always positive numbers), shifts the skin tones between yellow/brown and pink/magenta. You need to be careful as it is only necessary to move the curve a few points to make significant color shifts. Attached is the same image adjusted in LAB color. I touched on the magenta of the cheek and reduced the value of that point in the a channel about 3 points. The Magenta/red were reduced without affecting the yellow which is controlled by the b channel. I ended up at about the same place as the adjustment using selective color.

Oh yes, one more thing to mention. Adjusting white/black/neutral points is also a piece of cake in LAB because any point which is any shade of black thru white has a and b values equal to 0. This means clicking on any point and moving that point to 0 in both the a and b channels will neutralize that point.
Regards, Murray

Skin's not all that subjective. Sure, it can cover a whole host of hues and values, but there are some things it can't be, and that's always a good place to start.

Your first version, the one that "looked good" to you, is, in my opinion, the best of the lot, at least as a starting point. None of them are useable as is, but the version you corrected by eye has a decent magenta/yellow balance, though both are way too high. As is cyan.

How do we know? Well, it looks dark and heavy, which is a good clue. But the two spots on her face that qualify as bona fide highlights, the top left of her forehead and the tip of her nose, are reading 45M46Y with cyan in both cases coming in at more than half those values. Usually cyan is going to be around 1/5 to 1/3 of your M and Y. And for highlights, relative though they may be, it's safe to say that 45 is way high for M or Y. Knowing your impossible conditions is a good place to begin. Correct those and your remaining options will all be much closer to acceptable.

I moved the file into CMYK (using the default SWOP separation—if the destination space was CMYK I'd have given more attention to the conversion, but this was just one stop on the path) since there are two things that can be accomplished in CMYK that RGB and LAB can't duplicate. First, you can work with the black plate which is a great way to deal with shadow detail without affecting color. Those two highlight areas I mentioned also have 2 and 3 percent black in them, and the values go up from there, appearing in all the areas that normally would have only color information. (You don't want black creeping into skin tones before the halfway point.) The other CMYK move is to pull a sharp contrast curve in the magenta channel in luminosity mode. Another good trick to stretch out the darks and lights without messing up the overall color (recall that the M and Y were fairly well balanced to begin with.) After that it's just a bit of dithering to pull cyan into a reasonable range, at which point we move into the region of preference rather than science.

Here's my image after these moves:

http://edgework.tripod.com/samples/cmyk_moves.jpg

At this point our forehead highlight is reading 12C 29M 31Y which still might be a little cyan heavy, but we're in better shape for further moves.

Next I moved into LAB and ran Shadow/Highlight on the lightness channel, a fairly high amount with a narrow range, with the opacity pulled back to taste. Likewise a color enhance move pulling the anchors of the a and b channels 10 units each toward the center. This keeps the center point in place which avoids creating a cast, but serves to pull the colors further apart from each other. I reduced that curve's opacity as well. I also added a slight contrast move, then sharpened with the high-pass filter back in RGB.

http://edgework.tripod.com/samples/girls_skin_final.jpg

I chose not to assume that the background needed to be dead neutral, though, as Mistermonday pointed out, making it so is a breeze in LAB. Right now the values are 12C 17M 21Y, but I think the image is naturally meant to be warm.

With the exception of the cyan moves in CMYK, everything else focused on bringing out detail, which is often much more important, particularly when exact color is uncertain.

Hi edgework

Good job… but I’m not so sure about skin tone not being subjective. Your version, mine, and mistermonday are all within the limits and they look good to us but they are still slightly different. You have a slight magenta caste in the skin and it shows in the neutral background…I have a slight yellow… but very small on each count. This is what I mean by subjective.

For all we know the subject could have mixed heritage or a suntan or unusually pale skin… she could be flushed from heat or cold… None of these factors are we privy to.

Skin is always a guess… it can be and should be an educated one but it all comes down to what looks right to us.

Butch

...you have a slight magenta caste in the skin and it shows in the neutral background…I have a slight yellow… but very small on each count. This is what I mean by subjective.

As I pointed out, once I addressed the impossible conditions, we were no longer in the realm of science, but of personal perference. So yes, we're saying the same thing. I also specifically stated that I chose not to assume the background was neutral, that I felt the image should be a bit warm and the numbers of the background bore that out. Again, preference. Starting with an image that has so many problems (I'm assuming the original, which we did not get to see, was in fairly bad shape) you have to throw everything you have against the wall and be happy with whatever sticks.

I do hope you know edgework, I was not criticizing your good work at all…I was just making a subjective comment....:D

I do hope you know edgework, I was not criticizing your good work at all…I was just making a subjective comment....:D
No problem.

edge work...your pic is not coming up...

the first ...not the second

Hi Switters,

Do you have the original image, so we can see the starting point ?

I'm posting a jpeg from the DVD 'Creative Photographic Techniques for PS3' by Chris Orwig .... you can find it at Lynda.com - please delete it if it isn't allowed.

He does a video on there about skin, but also recommends correcting the image first, before starting. You may use levels or curves to do this, but the main point is, to get the image to a good starting point.

the first ...not the second
Odd. They work fine here. I even deleted my cache just to make sure I wasn't loading something old.

this is a good topic, someone should post a cleaner shot, maybe a photog here has a sample they would like to share.

there would be a lot of useful tips picked up here if so.

to the OP, i would linearize that image first, setting white and blackpoints and open it up a bunch, then concentrate on cloroizing it.