Skin tones with flash too... everything!

Snapper

I've been successfully using the theory in the Skin Tones tutorial for some time. However, when shooting snapshots at night with flash I often get results like this linked photo (flash bounced from white ceiling):

http://www.igmusic.co.nz/epix/20070607_0517.jpg

where the skin is too... everything!

The white of the t-shirt and the gray of the sweater indicate there's no obvious overall colour cast, but applying the Skin Tones theory to the nose for example, we have too much magenta, yet I can see green up near the hairline too, so I don't want to be adding more magenta to counter that. The overall look is that the skin is already far too yellow, even though there is more magenta than yellow.

I know, I know - trust the numbers! But my monitor is well-calibrated, and a good skin tone looks good on it.

It's as if I just want to dial-down the saturation of the reds and yellows, but that has the effect of making the skin kinda gray.

pathfinder

I am no expert at correcting color casts, so......

Was she sitting under a tungsten light? I see her as too yellow along her hairline, but too blue on her shirt sleeve....

Her face is way too dark, and very red, almost brown. She needs more cyan to kill the red, and she needs to be lighter as well.

This can be done with curves, or just go to Image> Adjustments>Variations and see if you don't like her better lighter, and with more cyan.

I hope that contortion was done with the Liquify filter. She looks like she is really in pain.

MarkR

Snapper wrote:

I've been successfully using the theory in the Skin Tones tutorial for some time. However, when shooting snapshots at night with flash I often get results like this linked photo (flash bounced from white ceiling):

http://www.igmusic.co.nz/epix/20070607_0517.jpg

where the skin is too... everything!

The white of the t-shirt and the gray of the sweater indicate there's no obvious overall colour cast, but applying the Skin Tones theory to the nose for example, we have too much magenta, yet I can see green up near the hairline too, so I don't want to be adding more magenta to counter that. The overall look is that the skin is already far too yellow, even though there is more magenta than yellow.

I know, I know - trust the numbers! But my monitor is well-calibrated, and a good skin tone looks good on it.

It's as if I just want to dial-down the saturation of the reds and yellows, but that has the effect of making the skin kinda gray.

Another option, assuming you're using Photoshop: Edit ==> Assign Profile, create a false profile with a 1.4 or 1.2 gamma. I used the 1.4 Apple RGB tat came on the Professional Photoshop 5e disc. Then Edit ==> Convert to Profile, convert it back to your working space.

You should then have more room to apply curves, etc. to tweak your final photo.

EDIT: Here's what I come up with, (hope you don't mind) after using a false profile (Apple 1.4 Gamma) to lighten the entire image, then using some simple curves to try to get the skin tones in l line:

edgework

The Hobbyist wrote:

Another option, assuming you're using Photoshop: Edit ==> Assign Profile, create a false profile with a 1.4 or 1.2 gamma. I used the 1.4 Apple RGB tat came on the Professional Photoshop 5e disc. Then Edit ==> Convert to Profile, convert it back to your working space.

You should then have more room to apply curves, etc. to tweak your final photo.

EDIT: Here's what I come up with, (hope you don't mind) after using a false profile (Apple 1.4 Gamma) to lighten the entire image, then using some simple curves to try to get the skin tones in l line:

I don't think the image is damaged enough to call for that drastic an approach.

When skin tones get weirdly out of whack, they seldom are too cool. It's almost always reds and yellows at work. A Selective Color layer can give you a good head start before moving on to more targeted approaches. In this one I set magenta to -14 and yellow to +14 for the Reds, and the yellows were set to +18 Magenta and -16 Yellow. The numbers seem large but they didn't reallly do all that much, just pushed the extremes back towards the center. Then I used a fake black plate as a layer mask for Shadow/Highlight. I tried a luminosity master curve in RGB but, predictably, it had kind of a muddy effect, so I took it into CMYK and used a magent curve in Luminosity mode with much better results.

MarkR

edgework wrote:

I don't think the image is damaged enough to call for that drastic an approach.

I wasn't under the impression that false profiles were for badly damaged photos. In this particular case it was a very quick way (once you have the profiles pre-setup, that is) to do two things: 1) lighten a dark photo, where most of the information was in the bottom quarter-tone, and 2) reduce the "too much" colors into a colorspace that gives more room for maneuvering.

I will say that on second look, I probably have too much blue in my version as a side effect of trying to kill some of the yellow (I doubt they have a purple couch!)-- dragging the blue curve down about 10 points in the midtones should kill that, I think.

jjbong

edgework wrote:

Then I used a fake black plate as a layer mask for Shadow/Highlight.

Could you elaborate a bit on this technique (or refer me to a post on it)?

edgework

jjbong wrote:

Could you elaborate a bit on this technique (or refer me to a post on it)?

Make a copy of the image and convert to cmyk. Pull the black plate back to your original as an alpha channel or else activate it as a selection and drag the live selection back. Now you can turn it into a selection and/or a layer mask (invert it first), targeting only the areas that Photoshop would define as black, were it an actual cmyk image. Allows a more refined approach to shadow detail than simply using RGB curves in luminosity mode, which can have a kind of bludgeon effect.

Snapper

edgework wrote:

I tried a luminosity master curve in RGB but, predictably, it had kind of a muddy effect, so I took it into CMYK and used a magenta curve in Luminosity mode with much better results.

Interesting. So by doing this you reduced the magenta level without increasing the green? I assumed that's what you meant so I went to CMYK and put a curves layer in Luminosity mode, pulled the magenta channel down, and it improved things immensely. If it's not what you meant, well it worked anyway!

pathfinder wrote:

I hope that contortion was done with the Liquify filter. She looks like she is really in pain.

Nope, no Liquify! It's genetic!

edgework

Snapper wrote:

Interesting. So by doing this you reduced the magenta level without increasing the green? I assumed that's what you meant so I went to CMYK and put a curves layer in Luminosity mode, pulled the magenta channel down, and it improved things immensely. If it's not what you meant, well it worked anyway!

It's not exactly intuitive what it does. Given that it's in luminosity mode, it, theoretically, has no effect on hue and saturation whatsoever, only value. In cmyk, I find that flat face tones are usually the result of a flat magenta plate. Using a magenta curve in luminosity mode has, as you discovered, a dramatic effect on contrast and detail. It's not specifically targeting magenta, per se, since it's not targeting any color directly, but the contrast move is working through the profile of the magenta plate, as a kind of layer mask, although it would be a mask that is dynamic, being affected along with the whole image as the curve steepens. That's as near as I can figure, but I don't really understand it fully. Then again, I don't need to understand electrons in copper wires to work a light switch. And this works.

edgework

The Hobbyist wrote:

I wasn't under the impression that false profiles were for badly damaged photos. In this particular case it was a very quick way (once you have the profiles pre-setup, that is) to do two things: 1) lighten a dark photo, where most of the information was in the bottom quarter-tone, and 2) reduce the "too much" colors into a colorspace that gives more room for maneuvering.

I will say that on second look, I probably have too much blue in my version as a side effect of trying to kill some of the yellow (I doubt they have a purple couch!)-- dragging the blue curve down about 10 points in the midtones should kill that, I think.

You're right. It's just that I've always resorted to a profile correction when more traditional approaches didn't give me the results I wanted. I think your move went too far, which is the problem with profile conversions: they're a brute force approach, and that's not always required.

MarkR

edgework wrote:

You're right. It's just that I've always resorted to a profile correction when more traditional approaches didn't give me the results I wanted. I think your move went too far, which is the problem with profile conversions: they're a brute force approach, and that's not always required.

Funny, usually my wife's the one accusing me of going too far ...

The nice thing about going too far in Photoshop is that you can usually dial back your changes

I made some assumptions about the girl's skin that might not be accurate. (Aside from the blue cast that I mentioned earlier. That's just me writing bad curves.) I assume(d) she is a caucasian/ fair skinned young lady. It's entirely possible that her skin is more bronzed/tanned (yellow) -- in your version the face is consistantly 10-20 points more yellow than mine.

Here's my latest correction, which corrects the blue cast & hopefully adds a little contrast.

Snapper

Pathfinder & Hobbyist - both of you have made vast improvements to my original. Yes, she's fair-skinned, tending towards the yellow. Hobbyist - did you use a false profile again in that last one? --Ian

MarkR

Snapper wrote:

Pathfinder & Hobbyist - both of you have made vast improvements to my original. Yes, she's fair-skinned, tending towards the yellow. Hobbyist - did you use a false profile again in that last one? --Ian

I used a false profile to lighten the image and reduce the color saturation (Apple 1.4 -- if you like it darker, Apple's default gamma is 1.8, which is still lighter than Adobe/sRGB) and applied curves from there.

I did play with plate blending a little bit (Green into the Blue in Lighten mode, 45% opacity) which helped kill some of the strong yellow.

The rest is just by-the-numbers curves. (And it's possible/probable that if I could write curves better, the plate blending would have been unnecessary.)

jjbong

edgework wrote:

Make a copy of the image and convert to cmyk. Pull the black plate back to your original as an alpha channel or else activate it as a selection and drag the live selection back. Now you can turn it into a selection and/or a layer mask (invert it first), targeting only the areas that Photoshop would define as black, were it an actual cmyk image. Allows a more refined approach to shadow detail than simply using RGB curves in luminosity mode, which can have a kind of bludgeon effect.

Amazingly powerful technique. I tried it on a few images using both Shadow/highlight (with a black inverted channel as a mask) and using Dan's overlay technique. The results were quite good, and the mask kept them right where you want.

I'm working on a summary of PP5E Shadow/highlight chapter, and I'll incorporate it. Thanks.

edgework

jjbong wrote:

Amazingly powerful technique. I tried it on a few images using both Shadow/highlight (with a black inverted channel as a mask) and using Dan's overlay technique. The results were quite good, and the mask kept them right where you want.

I'm working on a summary of PP5E Shadow/highlight chapter, and I'll incorporate it. Thanks.

I mentioned it in my Chapter 5 summary, which fit since it was all about the black plate anyway; but it's one of those simple ideas that has all kinds of potential particularly in workflows that never venture into cmyk. I'd love to see the ideas developed.