Hi there, I was wondering if anyone has any good tips on how to get the skin tone looking a bit better on this image.

I tend to stick in a fairly fierce S-curve in Photoshop to get this style of image (ie highlight fairly blown out and shadows deeper), however the downside is that there is always a yellowish tone to the skin. I have tried to ramp up the blue channel slightly, however that can lead to further problems. Does anyone have a suggestion for a good technique for correcting skintones or know of a good tutorial. It would be most appreciated, thanks alot for your time and attention

Freddy

I tried adjustment -> hue/saturation -> lowering the yellows saturation to -40

I used a selective color adjustment layer to add cyan to the yellows. I also decreased the contrast.

I quite liked Yings version, one thing though I am trying to affect just the skin area and not desaturate the image of colour elsewhere (eg the t-shirts) Is there anyway of isolating the skin quickly (quicker than a Quickmask anyway)

I find making the adjustment to taste and using a mask and painting in and out the detail much more efficient. It's rough, very rough, but just to give the idea

I used a selective color adjustment layer to add cyan to the yellows. I also decreased the contrast.

You're correct in seeing that the problem is in deficient cyan. However, using selective color to pump missing tone through another plate (in this case, building cyan information that matches the yellow profile) will simply pull your tones towards duotone quality. (Also, since yellow/blue information contains most of the grunge, using that color to add Cyan will make the skin extremely dirty.)

What's needed is to get Photoshop to build a new plate for you. Here's a trick from Dan Margulis:

Dupe the image and convert the duplication to cmyk. All things being equal, the screen representation should look the same for both images. Note, however, the difference in the Cyan plate and the Red plate. In theory, cyan ink reproduces the information in the Red channel (minus shadow detail siphoned off into the black plate), but the dot-gain compensation kills detail, particularly in the 25% - 75% range. Note also that the image we're starting with has hot Hot HOT shadows. So there's not all that much cyan information to move over to the black plate anyway.

Back to the RGB image, go the the Red plate and select All and copy.

Return to the CMYK image, go to the Cyan plate, select all and paste. You've now effectively boosted the Cyan channel with bona fide cyan information rather than tainting the Magenta or yellow plates.

Convert the CMYK image back to RGB and note the difference.

I actually repeated this process a second time, blending the final result into the original at about 80%, then used a slight Selective Color move to pull a little magenta and add yellow in the reds, and pull yellow and add magenta in the yellows to bring the tones more in balance. You can see that the shadows have now lost their heat and the rest of the skin tones are in acceptable ranges.

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

I tend to stick in a fairly fierce S-curve in Photoshop......

Seems that you are working in RGB mode.

RGB is great for your monitor to work on, but causes all sorts of problems when adjusting images.

When you lighten part of the image you are adjusting three channels (R, G, B) at the same time. The actual luminosity (lightness) that you see is a weighted average of these three. Problem is, when you raise them it's easy for one or two of them to max out (Red first, then Green), that's why skin turns yellow.
Full Red, full Green, less Blue is the recipe for a nice yellow tone.

Wouldn't it be nice if there was a way to adjust just the lightness and leave the colours alone?
Well, there is!
Convert to LAB mode (Image>Mode>LAB Color) - L is Luminosity and A and B are the colour information. Now run your S-curve only on the L channel. You change only the Luminosity and leave the colours alone.
When satisfied you can convert back to RGB.

Rô

Hello byRo

I was just wondering, will making a duplicate and setting its blendmode to luminosity then adjusting with levels or curves in RGB, give close results to lab mode?

Butch

That works pretty good too, Daviskw.

The results aren't exactly the same, but it's a lot better than adjusting RGB directly.

Rô

I'm not 100% sure on what you are looking for, so I just eyeballed it and did what I would have liked.

I just went to color range, and selected her skin color (which matched the child's), then I went into hue/saturation and shifted the hue left a little, and desaturated it.

Then I went to Selective Color, and took out some red from the black.

And lastly I went to Color Balance, put in a little Cyan in midtones, set to Preserve Luminosity.

I just tried a little of everything, basically.

I vote for just changing the curve adjustment layer to luminosity blend mode. ;)

do we get to vote?

If you like some of the increased color but not all of it, dup your curve adjustment layer, change one of them to color and the other luminosity. reduce the opacity of the curve layer that is using color blend mode to taste.

My vote goes to Roger's suggestion which is essentially the same as Byro's suggestion. Byro converts it to Lab and does the S-curve on the Lightness Channel whereas Roger suggests setting the Blend Mode of the Curves Adjustment Layer to Luminosity thereby also nullifying any effect to the colour. Either of these suggestions would have avoided the yellow problem in the first place. While all the suggestions in this thread are very creative ( and I especially enjoyed Edgedwork's - thanks for such a clear explanation Edge - I went and tried it immediately and it is a tip I shall note down for future reference) none of them are necessary because the colour cast shouldn't have been a problem in the first place. Why don't you post the original before the S-curve and we can have a shot at that and see if we can get the same contrast without affecting the colour?

Sincerely Syd

I usually set my layer mode to something other than Normal. Color and Hue do not effect luminosity and Luminosity doesn't effect color.

http://forums.dpreview.com/forums/read.asp?forum=1006&message=18831452

One way I fix skin color is by duplicating the layer in Color mode and then run Replace Color. I replace it using skin color charts (try Google Images).

Examples of my Replace Color action

http://forums.dpreview.com/forums/read.asp?forum=1006&message=18951902

just tried apply image very quickly to try and fix the toning in the channels.

Working with skin tones in LAB is really easy if one starts with the basic Caucasian skin color of 80 -20 - 20 (L - A - B). this issue bcomes where to apply that metric. In this case, i selected the baby's cheek. First picking the black and white points using curves and then adjusting only the skin tones as above gave the result in the thumbnail.

Look pretty natural to me ... what do you think?

Stosh

hello.

lovely pic.

1) autolavels then fade to about 10%
2) desaturate a bit
3) sharpen
4) filled a layer with blue and overlayed about 5%

cheers

.....the basic Caucasian skin color of 80 -20 - 20 (L - A - B). I prefer to use relative numbers for A and B instead, in this case A=B, and let the circumstances decide the value to be used.

Actually, I was just going to comment that I use B with 4 to 8 points more than A, but then I remembered that you said Caucasian so A=B is about right.
It's for Latin skin (like here in Brazil) that B is slightly higher (i.e. yellower).

Rô

I agree with you Rô, skin tones vary by race and even more so within the same photo. My decision is always - to which part of the face or body part to apply the metric. Averaging over a wider area seems to make sense, but the fine tuning always resolves to subjective tweaking. The cook book methods only take you to the general area, the rest is up to l'ouvre.

Stosh

I like the results from everyone it really has been informative. I like to try to keep the background white. So layer fills for me don't really work. The LAB colour technique certainly sounds interesting I shall now have a go at that. One question however, how is an image affected when it is swapped between different modes (eg RGB to LAB or CMYK?) am I degrading my image each time I convert it (I was some how under the impression that this was the case!)

Oh something else that effects my workflow is that I shoot all pictures using RAW, which is then converted using a raw convertor program such as CaptureOne or I am currently experimenting with Lightroom. This means that the inital conversion and tweeking of the pictures curves is carried out before I play around in Photoshop (therefore I guess the "yellow" damage has already been done, eg As far as I am aware I cannot work in LAB mode) so I guess I then have to fix it in Photoshop. Sounds to me rather like I am doing something wrong here! This is not such a good workflow, or is it normal?

Hi FF:

No need to worry about file Degradation converting back and forth to/from LAB in Photoshop. Just be certain the use Image>Mode>Convert to Profile command and select LAB, CMYK, sRGB or whatever, from the listing. Many studies have proven that the conversion is lossless, having converted back and forth thousands of times in the testing. A good primer onLAB is "Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace" by Dan Margulis $20 at Amazon.com.

So far as workflow from RAW, I would set the software preferences, in whichever is your program choice, to do a minimum of enhancement and adjustment and depend upon your own post processing skills to yield the desired results. Control is everything and this workflow will allow you the most flexibility.

Just one man's opinion.

Stosh

a quick guide to get nice skin tone is tou do this:

- copy pivture to new layer
-go to blur-avarage
-invert the avarage colour
-change blending to screen
-adjust opacity

a failry quick and nice way to do an easy adjustment! but it's more of quick lazy one, not that correct one :)


Gerry