Using Multiple Anchor Points and Correcting “Gray” Skin

Posted: June 30th, 2009 | Author: admin | Filed under: Uncategorized | Tags: cmyk, color correction, numbers, retouching, rgb, skin, skin color, skin tones | 5 Comments »

A Little Color Theory

Up to now, we’ve been pretty focused on skin the context of CMYK colors. Let’s switch it up for a minute and talk about skin as expressed in RGB values.

You already know that when skin is “correct,” Y>M>C (yellow is the highest; cyan is the lowest.).

How can you express that in RGB? You just turn everything on its head: swap each color for its opposite, and flip the mathematical terms.  The opposite of yellow is blue; the opposite of magenta is green; and the opposite of cyan is red. Soooo …

Y>M>C   or C<M<Y

AND

B<G<R   or R>G>B (red is the highest; blue is the lowest. ALWAYS.)

Why does that matter? Think about what else you know about RGB. Even if you’re not familiar with color theory, you should know that when a color is neutral (black, white, or neutral gray), R=G=B. Anytime you bring the R, G, and B values closer together, your color will approach gray. This is particularly important to understand in the context of skin. Since red will always be the highest value for skin, anytime you reduce red (add cyan), you are bringing the R value closer to the G and B values — which brings the skin closer to gray.* Sometimes this is necessary, especially when skin is in the shadows (shadows should always look a little more grayish than the rest of the skin, right?).

* Similarly, since blue is always the lowest value, anytime you increase blue, you are also bringing the values closer together (Blue closer to Red and Green) – this also has the effect of making skin more gray. Today, however, we’re going to stay focused on red/cyan.

With that in mind, let’s look at a couple of methods for correcting skin. One is tried-and-true: curves (groan).  The other uses still curves but starts off a little differently.

Method #1: Correct Color Using Curves Only

Here we have a picture from photographer/blogger/instructor Jodi Friedman (of MCP Actions fame). The little girl is super cute, but the color here looks just a little off — the skin has a bit of a grayish quality, probably because she’s standing in the shade.  Let’s take a few readings and see what we have:

Sample Point #1 (midtone: chin): These colors look fine. The cyan is a little less than 1/3 the value of the yellow. It could stand to be a bit lower, but it’s nothing to get worked up over.

Sample Point #2 (midtone: shoulder): Here the yellow and magenta are equal, which is fine, but the cyan is almost 1/2 their value. The cyan needs to be reduced here.

Sample Point #3 (shadow: neck): The yellow is 11 points higher than the magenta (which is too much), and the cyan is almost 1/2 the value of yellow. Cyan needs to be reduced, and magenta either needs to be increased, or yellow needs to be decreased.

The common thread in all three points is that the cyan is too high (it’s fine in point #1, but we’re likely to see an improvement by bringing it down to 1/5-1/4 the yellow value).In short, we’re seeing skin that is grayer than it should be.

Based on the tutorials of past weeks, you should know that a quick way to bring down cyan is to increase red, so let’s do that.  I’m just pulling right up from the middle:

Well, the skin certainly doesn’t look gray, but now the whole photo has taken on a fairly heavy reddish tint (the hair especially looks awfully red). We’ll add a Red Channel layer mask (Image>Apply Image>Source:Red Channel).

Much better, I think. Let’s check the numbers:

Sample Point #1 (midtone: chin): These colors are close. The cyan is a little less than 1/6 the value of the yellow (a little low), and the magenta is 1 point higher than the yellow (a little high).

Sample Point #2 (midtone: shoulder): The cyan is 1/4 the value of the yellow, which is great. The magenta is 2 points higher than the yellow, though (a little too high).

Sample Point #3 (shadow: neck): The yellow looks better here — it’s 7 points higher than the magenta.  The cyan is a little less than 1/3 the value of the yellow, which is fine for a shadow.

In short, the magenta is too high in the midtones, but it’s just right in the shadows. What to do? How can we fix the magenta on the chin and shoulder without messing it up on the neck? Simple: we add an anchor point to protect (to “anchor”) the color on the neck, and then we add additional anchor points to correct the color values on the chin and shoulder.

Take a look:

The anchor on this curve repersents the sample point on the neck. I added this point to the curve by hovering over the sample point with my eyedropper tool and then pressing CTRL+Shift+Click. This adds one point to every channel curve, and it represents the R, G, and B values for that color sample.

Now I’m going to add a second anchor point to the curve (again, pressing CTRL+Shift+Click with my eyedropper tool over the sample point):

Now, when I reduce the magenta at this point, it will not affect the magenta at Anchor Point #1 (Sample Point #1). Take a look:

I’m going to do the same thing on the red channel. The cyan looks good for Points #2 and #3, but it’s a little low on Point #1. We already have anchor points set on each curve to represent Points #2 and #3 (remember, when we added the points to the Green curve, we also added corresponding points on the Red and Blue curves). Let’s go add an anchor point to represent Point #1, and then we’ll pull that down a bit:

Here’s where we stand:

The numbers look good. What about the pic?

It looks fine. Not great, but an improvement. There’s still a bit of a cast coming onto her forearm from the skirt, and overall, it looks a little too red to me (even though the numbers are correct). We could keep tweaking, could use masking and multiple layers, or …

Sometimes — not always but sometimes — there’s an easier way.

Method #2: Correct Color Using an HSL Adjustment Layer to Reduce Cyan (Followed by Curves)

Remember when we talked about how high cyan levels means that skin tones are approaching gray (you were wondering when I was gonna get back to that, right?)? Well, turns out there’s another way to reduce the cyan OTHER than pulling up on your red curve.  Think about what you would do if you wanted to pull your overall picture TOWARDS gray. You’d reduce saturation, right? Well if you want to move your image AWAY from gray, bumping up saturation can be an easy way to do it. There are many ways to increase your saturation (I’m sure you’ve read a dozen different ways to “pop” your color), but we’re just gonna stick with the simplest and crudest method here (though I recommend using a more advanced method, such as LAB mode, if you are familiar with that sort of thing).

Back to the original image:

All I’m gonna do is pull the saturation up to +27:

Already this is a big improvement.  Compare this to the first step in Method #1, where we simply pulled up on the red curve — this looks MUCH better. Let’s check the numbers:

Look at what has happened — across the board, cyan has DECREASED, and both yellow and (to a slightly lesser extent) magenta have INCREASED.

Sample Point #1 (midtone: chin): The relationship between yellow and magenta looks great. The cyan is a little low here (nearly 1/8 the value of yellow), but overall, this looks pretty good.

Sample Point #2 (midtone: shoulder): These colors look great – yellow is just slightly higher than magenta, and cyan comes in at 1/3 the value of yellow. Cyan could go a little lower or stay as is.

Sample Point #3 (shadow: neck): Cyan is coming in at 1/3 the value of yellow – which is perfect. Magenta is 15 points lower than yellow, which is too low.

Here’s the plan:

1)  I’m going to add three anchors – one representing each sample point – on the R, G, and B curves (do this by CTRL+Shift+Click on the sample points with the eyedropper tool).

2) Then I want to increase the cyan (pull down the red) for sample point #1 — without changing the cyan values for points #2 or #3.

3) Then I want to reduce the yellow (pull up the blue) for sample point #2 – without changing the yellow value for points #1 or #2.

Let’s get to work…

First I add all my sample points to the curves. Here’s the red curve (Sample Point #1 is selected):

I’m pulling down on Sample Point #1 just a bit, and …

My curve looks a little goofy, but the numbers look great! Let’s bump up the magenta (pull down green) at Sample Point #3:

And the result:

I’m not saying it’s perfect – there’s still that magenta cast coming off the skirt (on the forearm), and there’s some excess yellow in the shadows. That stuff can be fixed with a little extra time and some masking. But for THIS picture, bumping the saturation before starting with curves is quicker AND yields (I think) nicer results than working with curves alone.

Take a look (Top: Original; Middle: All Curves; Bottom: Saturation + Curves)

And some side by sides:

Original vs. All Curves Adjustment

Original vs. Saturation + Curves Adjustment

Don’t like the extra saturation in the rest of the pic? You can always mask it off …

Bear in mind that this is just one method to add to your arsenal. Sometimes it’s better to work all with curves. Sometimes it’s better to work with various other adjustment layers (we’ll get to those).  Sometimes — particularly when the yellow/magenta ratios already look good — doing a saturation bump will solve all your skin color problems right off the bat.

Here’s the take away:

* Anytime you reduce the amount of red in skin, you bring the skin color closer to neutral gray.

* As far as skin is concerned, the saturation bump widens the gap between cyan and yellow/magenta.  Thus, you should only use this method for images where the skin has an EXCESS of cyan.  If the skin has too little cyan (too much red), the saturation bump will only make the skin look worse (because it takes the cyan even lower relative to the yellow/magenta).

* The saturation bump will NOT work if for some reason your skin color is so far off that Red is lower than Green and Blue (viewing skin color in terms of RGB) — this is almost never the case, but it’s something to consider.

That’s it for this week! Thanks again to Jodi for the sample picture.

---