Skin-tone question for great EZPrints

jfriend

I'm working on some skin tone color corrections for prints that will be made through smugmug. I find this smugmug help page very helpful when I'm in Photoshop and have access to CMYK values, but I have a couple issues that I'm looking for some other reference sources or help on.

First, I'm trying to color correct a group of images in the ACR RAW editor where I only know how to see RGB values, not the CMYK values discussed in the smugmug help page. Does anyone know of a similar document to the smugmug help file that talks about how to interpret RGB values for skin color correction or how to do good looking skin tone from Adobe Camera RAW?

Second, when I load the image below into CS2 and look at the CMYK values like the smugmug help page says, I have C, M and Y values that are relatively OK versus each other (C is the lowest, then M, then Y is the highest), but my M and Y values are significantly higher than those discussed in this help page. Some representative values from the mom's face are:

Center of forehead: C:20%, M:49%, Y:60%
Darker area on her right cheek: C:12%, M:72%, Y:90%
Medium tone area on her right cheek: C:26%, M:63%, Y:73%
Medium tone area on her left cheek: C:25%, M:58%, Y:68%

If I lower the M and Y values down to the range discussed in the help page using a technique discussed in that page, the little girl's face continues to look OK, but the mom's skin looks really gray and unpleasing on my test prints at home.

So, are these higher values of M and Y OK for this image and will print fine on EZPrints? Or do I need to do something different? And, any ideas how to do skin tone color correction in ACR? I've followed the other active thread about skin tone

Here's one of the images I'm working on (the one the CMYK values above were taken from). Larger version is here.



Thanks for any help you all can provide.

DavidTO

Not sure how much help it is at $500, but Capture One Pro will allow you to view CMYK values in RAW conversion.

Andy

Hi John,

I convert from RAW and then do any fine-tuning in Photoshop. No, there's not a SmugMug help page on RAW, with CMYK, becuase AFAIK only C1 allows it. Wish I had a better answer for you on that.

I think you've done a fine job on the editing - it's really personal preference at this point - but IMO - the image (indoor light) is very warm, so I cooled it down a bit, just quickly applied a cooling filter to the image, and then masked away 50% of the cooling on the skin. Both images would print up fine on ezprints via SmugMug, selecting true color.

jfriend's edit on your left, mine on your right


All the best,

jfriend

Good to know

Andy wrote:

Hi John,

I convert from RAW and then do any fine-tuning in Photoshop. No, there's not a SmugMug help page on RAW, with CMYK, becuase AFAIK only C1 allows it. Wish I had a better answer for you on that.

I think you've done a fine job on the editing - it's really personal preference at this point - but IMO - the image (indoor light) is very warm, so I cooled it down a bit, just quickly applied a cooling filter to the image, and then masked away 50% of the cooling on the skin. Both images would print up fine on ezprints via SmugMug, selecting true color.

Thanks - good to know it should print fine as it is.

I think you might have misunderstood my first question. I was interested in a help source that guides you on adjusting for a pleasing skin tone when all you have is RGB values (because that's all most RAW converters have and all my RAW converter, ACR, has). If I had CMYK values there, the current document would work just fine.

Yeah, I guess this is a text-book example of mixed illumination. The foreground is lit with a mixture of bounced flash, direct flash and room light (SB-800 pointed up at 45 degrees with the white card out for a little direct flash illumination). That probably gives me three competing lighting colors (one from the flash directly bounced off the white card, one from the color of the bounce off the beige ceiling and walls and the ambient incandescent light).

I hadn't though about cooling the background. I'll play with that some, but I think you're right, that's just a question of general photo aesthetics, not about pleasing skin tone.

Andy

jfriend wrote:

Thanks - good to know it should print fine as it is.

I think you might have misunderstood my first question. I was interested in a help source that guides you on adjusting for a pleasing skin tone when all you have is RGB values (because that's all most RAW converters have and all my RAW converter, ACR ,has).

I'll look into that!
BTW I did say it would print up fine (true color) but - this type of shot will print better (anywhere) if it were a little cooler IMO - the cooling down has also the effect of lowering the saturation, and with the indoor / mixed lighting, the saturation is up there...

rutt

A quick fix is to use Selective Color and add a little cyan (10% maybe) to both red and yellow. I tried this and it looks a lot more natural.

A much less quick fix, but worth it if you want to print this, is to take it through the Dan Margulis portrait workflow. Before doing that, though, use RGB curves or selective color as above to cool a bit. The advantage of doing this with RGB curves is that you can cool without darkening too much.

If you take this through much of a workflow, you'll want to surface blur the woman's skin but not the little girl's. Use surface blur on a layer and use blending options to restrict to the positive sides of the A and B channels. Than use a layer mask to further restrict to just the woman's face. You'll be able to use a very sloppy selection for this becase the blending options you just used will do most of the work of excluding hair, clothes, background, etc.

jfriend

I might try the Margulis portrait workflow

rutt wrote:

A quick fix is to use Selective Color and add a little cyan (10% maybe) to both red and yellow. I tried this and it looks a lot more natural.

A much less quick fix, but worth it if you want to print this, is to take it through the Dan Margulis portrait workflow. Before doing that, though, use RGB curves or selective color as above to cool a bit. The advantage of doing this with RGB curves is that you can cool without darkening too much.

If you take this through much of a workflow, you'll want to surface blur the woman's skin but not the little girl's. Use surface blur on a layer and use blending options to restrict to the positive sides of the A and B channels. Than use a layer mask to further restrict to just the woman's face. You'll be able to use a very sloppy selection for this becase the blending options you just used will do most of the work of excluding hair, clothes, background, etc.

I've read Dan's portrait workflow and read all your postings on the topic, but never actually tried one myself. I may give it a spin on this one. I'll post the results if I get time to do it.

rutt

I played around with this image trying to make a real improvement. It's harder than it looks at first. I'm not all that happy with my results, but it's time for me to stop. So, I'd like to share what I learned and maybe someone will find that useful. Maybe someone will do better than I did. I haven't set a very high bar.

Before I start, a couple of opservations about the image. What's going on in this image that makes it hard is the mixture of direct flash (direct enough to cause flash reflections on the woman's face), tungsten which is a much warmer light, and makeup on the woman's face. The makeup reflects light in a totally different way than the skin of the little girl. Next time, I'd try to bounce more and maybe try to get a more even mixture of flash and ambient light on their faces. I don't really know what to say about the makeup, but it's an issue.

I started off by trying to even out the worst of the tungsten influence. I took the image to LAB and used these curves:



to get this:



I moved both the A and B curves toward the cool side. The B curve also flattens out though the skin tones to get rid if the yellow cast on the girl's and hn the hands for example.

At this point I decided that the woman's face was going to need special treatment, probably because of the makeup, but also because of her age and gender. So I used the B channel to make a quick mask. Here is the B channel after Image->Adjustments->Auto Levels:



And here it is after a curve, some blurring, and a very sloppy selection/invert/fill move:



The first thing I used this mask for was as a layer mask for a surface blur. Surface blur has it's own detail preserving features, but here, I want to keep iit away from the little girl's skin altogether.

Here's a before and after of the surface blur though the mask:




After this I converted back to RGB in order to get a green channel to use as an L channel to try to get some more depth into the girl's face. This is basically following Dan's recipe, but perhaps in this image it was a mistake. If I had the time I'd try without this particular move. Or perhaps I'd use my handy mask to keep it away from the woman's face, which it doesn't seem to help.

Now I used my mask again to apply separate LAB curves to the woman's face and to the rest of the image. My logic here is that I was fighting the makeup. I want to get some rosy cheeks into the little girl's face without making the woman too magenta. Or put the other way, I want the woman to have a nice color balance without turning the girl too yellow.

Here are the curves I used for the woman's face:



And here are the curves I used for the rest of the image:



For the woman's face, I cooled even more and also flattened the L curve a bit to make her skin seem smoother. This L flattening is another questionable move, I think. Perhaps if I'd used the mask to keep the luminosity blend away from her skin, I wouldn't have needed it. For the rest of the image, I just did what worked for the girl's face, which was trying to up the magenta side of the A curve to get rosy cheeks and generally attack the jaundice left over from the tungsten. I admit to adding a layer mask and limiting this move in several key spots, including the hands and the framed picture.

Normally, I try really hard to avoid this kind of dual treatment, but the skin tones are so different, I didn't know what else to do really.

Here is the image after application of these two sets of curves:



Now I was back into the recipe. I sharpened twice, convention L channel USM and high radius low amount to arrive here:



I was unhappy with the plugged shadow detail in the hair, but we know how to handle that in CMYK (see Professional Photoshop, the chapter called K is the Key. I steepened the K curve though the area of the hair and then used Selective Color: Black to remove the about 10% of cyan, magenta, and yellow from the blacks and reveal some shadow detail. While I was at it I sharpened the K channel to bring out accentuate the newfound detail in the blacks. All this is pretty standard shadow unplugging technique.

So here's what I got after all of this.



Was it worth it? Perhaps more as an exercise than anything else.

jfriend

It is a tough image

rutt wrote:

I played around with this image trying to make a real improvement. It's harder than it looks at first. I'm not all that happy with my results, but it's time for me to stop. So, I'd like to share what I learned and maybe someone will find that useful. Maybe someone will do better than I did. I haven't set a very high bar.

I don't have time to read and respond in detail right now (I will later tonight), but I agree with your conclusion and I thank you for posting your learnings. I played with this for 6 hours yesterday (more as a learning exercise than anything else) and took it through the Margulis workflow. What I concluded was that the woman's face just needed a whole different workflow than the girl and the girl's face was a lot easier to make look great. I didn't think about make-up as a core cause, but that makes some sense. I also spent a bunch of time removing reflections and skin shine which helped the overall look a lot, though it has nothing to do with the Margulis workflow.

More later.

rutt

jfriend wrote:

...and the girl's face was a lot easier to make look great. I didn't think about make-up as a core cause, but that makes some sense.

Makeup isn't the only reason why the woman's face needs a different treatment than the girl's; young skin is just beautiful the way it is and the woman's face needs a blur early on. The flash reflections on her face are probably the result of sweating through makeup. In my experience that really catches the reflections and here it explains why the little girl doesn't have this problem.

The reason for the yellow below the girl's neck is that the flash is direct, not bounced off the ceiling. So that part is lit by the ambient tungsten.

Now you know why I didn't include and of these group shots in the first portrait practice set.

jfriend

Anatomy of a retouch

rutt wrote:

Makeup isn't the only reason why the woman's face needs a different treatment than the girl's; young skin is just beautiful the way it is and the woman's face needs a blur early on. The flash reflections on her face are probably the result of sweating through makeup. In my experience that really catches the reflections and here it explains why the little girl doesn't have this problem.

OK, after countless hours fussing with this photo, I'm calling it a day and going to go with what I have. I've far passed that point of staring at an image so long that you just can't see the forest through the trees anymore.

Here's what I started with and what I ended up with (original on left, retouch on the right). The retouch steps I went through are below:



First off, I have several versions of this image that I tried the Margulis portrait workflow on. I was not successful with that workflow on this image, but it did finally teach me that workflow and I will write a separate message on that topic as I learned a lot that should be useful to others.

Here's what the retouch involved in outline form, then some more detail on each step below:

• RAW conversion in ACR
• Convert to LAB
• Do simple color enhancement of pushing in the ends of the A and B channels a little bit (I have forgotten exactly how much, but it wasn't much more than 10%)
• Slight contrast S-curve on the L channel.
• Convert back to RGB
• Mask out part of the LAB color enhancement on the woman's face (it was too red while the girls face was perfect)
• Remove glare on the chin with cloning
• Remove the skin shine on the woman with painting, blend mode and opacity
• Remove the major reflections in the woman's glasses with cloning
• Lighten the woman's face slightly using a gray overlay layer (I call it a burn/dodge layer)
• Bring out a bit of detail in the dark hair with a shadow adjustment on shadow/highlight.
• Mask out that shadow/highlight change to only apply to the hair (we don't want it to apply to the girl's dress for example as it just higlights dust on her dress)
• Apply a slight cooling filter to the background since the lighting and LAB changes made it way too warm (masking out everything but the background)
• Brighten the woman's teeth a bit using a Hue/Saturation adjustment layer and a mask (they became too yellow after the LAB adjustments)
• Do a Local Contrast Enhancement using Margulis' USM techniques
• Mask the contrast enhancement from most of the woman's face because it accentuates some detail we don't want to see
• Sharpen with smart sharpen
• Mask the sharpening so it only applies to eyes and mouth, eye brows and a partial effect on the hair.
• Darken the foreground hands a small bit using a mid-tone curve adjustment because the hands were too bright (closer to the flash and too prominent)
• Mask that curve to only the lower part of the image.

Here's the layers palette for the finished image with more detailed steps below:


Here's some more detail on the various steps:

RAW adjustments: Mostly what I was trying to do here was to make sure I preserved as much highlight detail as possible. There are a few blown out places on the woman's face and I wanted to make sure I preserved as much around those as possible even if my image came out of the RAW editor a little darker than needed. I left the default ACR curve. If I were to do this again, I would have chosen the straight curve because that would have left me with a little more shadow detail before going into CS2. But, once I got the LAB step the way I liked it, I didn't want to redo the RAW step so I left it the way it was. This was a suggestion by Rutt earlier which I agree with. If you are going to adjust contrast yourself in CS2, don't take away options in the RAW editor by apply a contrast curve there. You just throw away options for later.

LAB color enhancement: The LAB changes were just simple LAB adjustments of pushing in the ends of the A and B channels to drive steeper curves. I wasn't looking for a lot of difference here so I went modest (I think it was only about 10% on each end, but I don't rember exactly). I wanted to pull out a little more color contrast in the face and see if we could get a little better color in the woman's face. I found that no matter what I did, it was easy to get the girl's face to look great and hard to get the woman's face to look great and impossible to get them both to look good at the same time. I ended up deciding that I'd for a great face on the girl and decide later if I would do some separate corrections on the woman's face.

L Curve: I added a slight L-curve here for contrast and mid-tone brightness, but had to be wary of not losing even more shadow detail as the woman's hair was already darker than I wanted.

Mask out some of the LAB effect on the woman: After coming back to RGB there was just too much red in the woman's face so I create a mask and removed about 50% of the effect from her face with a soft-edged mask.

Remove Glare on Chin: A day or so went by and then I came back to the image. I realized at this point that I should have done some basic cosmetic correction before I started on color correction. But, I didn't want to have to redo the LAB enhancements again so I decided to do these cosmetic corrections on top of that. I experimented with a couple ways to get rid of the glare on the woman's chin. What I found worked best in this case was to clone from an area nearby onto an empty layer. Then put the layer into darken mode and dial the opacity down a small bit until I got a little bit of natural highlight on the chin (like there should be), but nohting that look like a reflection. This keeps the clone from looking "flat". I found that the combination of darken mode and a dialed-back opacity lets some of the natural skin texture and some of the natural tone variations peek through.

Remove Skin Shine: There was clearly some combination of direct flash, makeup and sweat that caused a bunch of skin shine on the woman that really detracts from her image. I had never actually done this before, but read a few Googled postings on the topic and came up with a technique that worked pretty easily. I created a new blank layer. I set the blend mode to darken. I then sampled a color near an area I want to fix with the eye dropper. I then selected a soft-edged brush and just painted that color over the area in question. Then, I dialed back the opacity to let some of the natural tone variation peek through to keep the replaced color from looking flat. The "darken" blend mode also lets the darker portions of the texture peek through and lets us only focus on eliminating the bright shine part. Before trying the darken blend mode, I also successfully used Blend If adjustments, but I realized that darken mode was easier and just as effective in this case. I used this technique for a bunch of different areas: on the woman - each cheek, nose and forehead, on the girl, nose. I was skeptical about it working on the large shiny area of the woman's forehead, but it worked great. The darken blend mode and dialed-back opacity keeps it looking quite natural. I did resample for a local color before each area to fix.

Here's all the areas I worked on on the woman's face and the result (note this is early in the workflow so skin tone isn't done yet - this is just showing skin shine and hotspot areas that were fixed):


Remove major reflections in woman's glasses: The reflections in the woman's glasses were pretty distracting. I experimented with a bunch of techniques here, but the only thing I could really find that worked was good old-fashioned cloning. One eye was harder than the other because the reflection crossed a feature bounday (pupil to eyelid). In that case, I made a selection of just the pupil area and cloned the reflection out of that, then inverted the selection and cloned the reflection out of the eyelid portion. There are still some star-like reflections in the glasses that I decided to leave there as they aren't that distracting and I wanted to spend my time elsewhere.

Lighten the woman's face slightly: At this point, I decided that the woman's face was little too dark and I'd seen that her skin tone looked more natural when the mid-tone brightness was raised a bit. I decided to use an overlay layer (I call it a burn dodge layer). Create a new layer, set to overlay blend mode, fill with neutral gray. Paint on the layer with a 15% opacity soft white brush to lighten any area you paint on. Light more with multiple strokes. I just made one pass with a very large and soft brush over the main part of the woman's face to bright it just a bit. I could have also used a masked curve.

Bring out some detail in the hair: At this point I was bugged by the lack of detail in both of their hair, but most specifically the woman. If I were to start this over again, I would have removed the default ACR curve because I think it's attempt to add some overall contrast caused some shadows to plug in the hair. But, I was feeling too far into it to start over so I figured I'd see what I could do with shadow/highlights. I made a new merged layer of everything so far and applied a gentle shadow adjustment (only about 10%). The girls hair responsed nicely at almost any adjustment, but the woman's hair started to get a shine and some streaks in it with significant adjustments (I don't know if these are coming from strands of gray hair of more flash sine). In any case, I settled for a little more detail, but not to the point where the shine started to show. I did go back to the RAW image at once time to pull up the shadow side of the curve to see how it looked and even took one partway through post processing derived from that, but in the end I decided that her hair is jet black and without adding auxiliary custom lighting we're just not going to see much detail in the hair without it looking unnatural. So, I settled for the middle ground of a little shadow adjustment. At this point I also got an opinion from my wife that the image where I was trying to preserve detail in the black hair looked unnatural so I abandoned that effort.

Mask out shadow/highlight change: I realized at this point that I really only wanted the shadow changes to apply to the hair. I had the face the way I wanted it and I didn't want to raise the girl's dress. So, I just masked out the S/H changes to only the hair area.

Cool the background: After discussing the image here earlier in this thread, I had come to the conclusion that the background was too warm. As Rutt and I have discussed earlier in this thread, this pic is a mixture of direct flash, bounced flash off a colored wall and incandescent lighting. The background is all lit with incandescent and after color correcting for the flash, it shows way too warm. It's warmness is also messing with our color perception of the people's skin. So, I used Andy's suggestion of a cooling filter and then masked the effect to only apply to the background. I was actually surprised how much this helped the image. My guess is that removing the overly-warm background removes color correction conflicts from the image and the brain can now do it's normal color correction perception processing on the whole picture. With the overly warm background the brain just isn't sure how to auto-color correct the image and gets confused. I experimented with leaving some of the cooling effect on the people, but found I ended up with M higher than Y in the skin and it didn't look right on my test prints at home. So, I confined the cooling effect to the background.

Brighten the woman's teeth: After everything we've done so far, I noticed the woman's teeth were a little yellow. I wanted to keep them looking natural but remove some of the yellow. To do so I create a hue/sat adjustment layer, dropped the saturation and raised the brightness and masked the effect to just her teeth. I then dialed back the opacity a bit when it seemed like maybe I had overdone it. It amazing how your first effort to fix something is almost always overdone and it's always great to use a technique that can dial back the opacity when you are done.

Local Contrast Enhancement with USM: Here I practiced Margulis' technique for USM with large radius. I followed his steps exactly. I set the amount really high (500) and played with the radius to find out which areas were affected at different radiuses. What I was hoping to do was to add a little more definition to certain facial features. I found that I liked a radius of 19. I then dialed down the amount until I liked it and tweaked the threshold up to 9. I optimized this setting for the girls face and found that I got accentuation of some detail I didn't want in the woman's face. So, I masked off part of the effect from the woman's face. This particular move gave some nice pop to the whites of the eyes and gave some facial features a little more definition.

Darken hands: At this point I thought I was pretty much done when it hit me that there was a little too much emphasis on the hands in the foreground. They were bright (perhaps because they were closer to the flash or just because of the contrast with the black dress) and I know some woman are sensitive about detail in their hands. So, I decided to lower the mid-tones a bit on that. I did so with a masked curve. I pin the dark tones with a point on the curve to they wouldn't change and then pulled down the mid-tones a small amount. I then masked out this change to only the lower part of the image. Since the darkest tones weren't changed by the image, the mask could be very sloppy.

Learnings

---

There were lots of learnings from this image. This was not a staged image. I found the two of them at Thanksgiving on the couch together and just asked them to smile for the camera. Here's what I would do different next time.

While Shooting:

• Shoot at a little larger aperture for this type of photo. This was shot at f/4, 1/60th and 52mm, but I could have used f/2.8 and still had enough depth of field. That would have given me more background blur and less powerful flash so hopefully less shine.
• Experiment with 100% bounce on the flash (without the white card up on the Nikon SB-800) so I don't get the reflections and shine in the skin.
• Buy a flash bracket (the CB Junior is on my holiday list) as this gives you better flash angles in the portrait orientation
• Get a softbox (this Lumiquest is on my holiday list) as this will also reduce shine and soften shadows
• When taking a portrait oriented shot like this, tip my flash to point to the ceiling. I had it at 45 degrees up from the camera which ends up bouncing off a side wall when in holding the camera in portrait mode.

On the Retouch:

• Remove the default ACR curve before going to CS2 as this gives you more shadow options in the retouch.
• Focus more on color cast removal in ACR than in getting a pleasing skin tone. I figured out 3-days into it that the starting image had a somewhat red cast to it (there's more red in the black dress than there should be). Margulis advises to remove all casts before color correcting/adjusting because they just get magnified in undesirable ways. I now think the image I started with out of ACR had a red cast to it and that's why the woman's face came out too red after the LAB adjustments. Margulis discusses how you can pick items in the photo that should be of a known tone. In this case her black dress certainly has no business having a lot more red in it than any other color. I could have probably just used the ACR eye dropper on that part of the image and started with a nice neutral image (at least for the flash illuminated portions).
• Fix skin blemishes early or late, not in the middle. I had a problem that I fixed skin blemishes in the middle of my workflow. That meant that if I later decided to change one of the earlier steps in my workflow, I had to redo the skin blemish fixes (because cloning or painting from sampled areas or colors will be off when you change what's underneath them. I actually had to redo the skin fixes three times due to this. I got good and fast at it and knew all the settings, but it was annoying. If I did them early, I would have only done them once. I supposed I could have also done them last (just before sharpening) when I was SURE the rest of the image was done, but even then I might still want to tweak something after seeing proof prints.
• Removed casts from the background like this one had earlier in the workflow. I think that the overly warm background was messing with the brain's perception of the skin color for a long while. In this case, it was a real easy cooling filter adjustment layer too.
• Experiment with surface blur for the woman's skin. In the end, I decided that as long as I kept the sharpening off the woman's skin, that it looked fine and I wanted to keep it natural, but I would like to experiment with using surface blur on her skin. Because this can't be done on an adjustment layer, this is probably something that's just like the cosmetic defects and you should do it early in the workflow so you won't get caught doing it over again if you want to change something else.
• Make my LAB adjustments in a separate copy and then paste the resulting merged layer in. For an image that you may spend a lot of time on, it's a real disadvantage that you lose your lab adjustment layers when you convert back to RGB. I do my LAB adjustments in multiple adjustment layers and can then freely tweak the A, B and L adjustments seprately, even adjusting opacity to taste. In the case of this image, I just converted to LAB, made my adjustments, flattened and converted back to RGB. I had then lost my LAB adjustments and couldn't easily get back to them to tweak them. An alternative which I'm building an action to do is to copy a merged copy of the image to a new image, make the LAB adjustments there, copy a merged copy back to your RGB image and keep the LAB adjustments around until you know you are truly done. In my case, I could have redone a few things from the RAW file, but still kept all my LAB adjustment layers without having to redo all that. This is only for images you are seriously going to mess with, but it would have saved me time in this case.
• This is kind of a duuuhh, but make sure you have a calibrated monitor. When I started working on this image, I did not have a calibrated monitor (I'd been getting by with Adobe Gamma). I bought the Gretag MacBeth system during this process and that helped a lot. I still don't have a perfect match with my printer, but they are not far off now.

That's all for now. I did learn a lot about Margulis' portrait retouch while trying that out on this photo (I did uses pieces of it) and I hope to write more on those learnings soon.