I believe that almost every photographer has been confronted by a photograph of one or more people, shot either under unknown lighting conditions or mixed lighting. Meaning that either a white-balance card (or other device) was not used, or that the lighting might be a combination of two or more of the following: daylight, flash, fluorescent, or incandescent light. The camera is usually set to Auto White Balance and the final result is an image that is not properly color balanced.
I have a quick method for dealing with this situation. Everything that follows also applies to film shooters, where again, the image is highly likely not to be properly color balanced (I’m assuming that if shot on film, the image is subsequently scanned and edited digitally.)
Over the years, I have observed and spoken to a number of photographers on this topic. Either using Photoshop or another editing program or a Raw converter, most of them do an adjustment to the color temperature, followed by adjustments in tint, color, and curves until things “look good.” Somehow, I always felt that this ad hoc method was inefficient, inaccurate, or not particularly appealing.
Therefore, I decided to find out what expert retouchers do to see if there might be a better and more effective way to address this issue. As anticipated, there are a number of interesting and more effective ways to achieve beautiful skin tones, and they usually correct the tones in the rest of the image at the same time. I describe below a personal variation on a particularly effective methodology.
The crux of the method
The percentages of yellow and magenta in any skin tone are approximately the same. This is an extremely important fact to remember. Please note that the word “any” is in italic letters. The reason is that this statement applies regardless of whether the person is Asian, African, Middle Eastern, Latino, or Caucasian. I have not yet found a skin tone that does not have approximately equal percentages of yellow and magenta. I do not say that they are exactly equal because the amount of yellow is usually slightly higher.
So let’s jump right in and open our image in Photoshop without any color- or white-balance corrections. Figure 1 is a portrait of my good friend and contributing editor to PHOTO Techniques Michael Reichmann. It was taken in late afternoon in Toronto, with a PhaseOne camera and a P45+ back in AutoWhite Balance mode.
The first step is to choose the Eyedropper tool and then open a new Curves adjustment layer. Make sure that the Info palette is visible in your monitor. Pick a point in the photograph where the skin tone is light, but not blown out. I usually choose the forehead or the chin, being careful not to pick a highlight.
While pressing the Shift key, click on the selected point. This puts a circle on that point and records the RGB reading in the Info palette. Now, go to the Info palette and click on the small triangle under the eyedropper (Figure 2). From the drop-down menu choose CMYK Color because this method is far simpler when working with CMY values rather than RGB values.
Remember that we are working with the CMY percentage values, not the values shown to the left of the percentage values. Also keep in mind the complementary colors: curves in Photoshop are marked Blue, Red, Green, and RGB. However, since reducing a primary color is equivalent to adding its complementary color, think of the color curves as Blue-Yellow, Red-Cyan, and Green- Magenta adjustments.
Going back to Michael’s portrait, the initial values in the image are: C: 44%; M: 46%; Y: 28%. So the first step is to click on the drop-down menu at the top of the curves window and select the blue curve (always start with Blue-Yellow adjustments). The easiest way to edit this curve is to select the Targeted Adjustment tool (TAT) by clicking in the small square with the hand symbol above the three eyedroppers in the curves window. Next click on top of the image in the circle marking the selected point, and while holding down the mouse button, move the mouse up or down until the Yellow value slightly exceeds the Magenta value. You can also edit by modifying the curve itself, but I find the TAT much easier to use. My first adjustment is shown in Figure 3.
Figure 4 shows the image after this first adjustment. It is immediately obvious that the color is much better than the original. In fact, it is so much better that there might be a temptation to stop right here, but the image can be made better with a little more work.
Here is where the second set of rules-of-thumb come in:
• For light Caucasian skin, the Cyan value should be approximately 1/4 of the Yellow and Magenta values.
• For ruddy skin, the Cyan value should be approximately 1/3 of the Yellow and Magenta values.
• For Asian or Brown skin the Cyan value falls somewhere between 1/3 and 1/2 of the other two.
• And for dark or black skin it is very close to 1/2 of the Yellow and Cyan values.
Note that these numbers are approximate; slight variations are quite acceptable at the end of the process.
So, going back to our image, my second step was to force the Cyan value (editing the Red curve) to 1/3 of the other two. This is the ruddy skin rule of thumb. Doing this unfortunately changed the Magenta value, and the image looked terrible. So I had to adjust the Magenta value to be close to the Yellow value, but this changed the other two values,so I had to adjust the Cyan again and then the Yellow again.
Once you get the hang of it, you can do several iterations very quickly until the numbers converge around the proper values. Once this has happened, you can do a final visual (small) tweak at the end. My final tweak in this case also includes a slight increase in brightness via a small adjustment in the combined RGB curve.
Figure 5 shows the final image. I hope magazine reproduction can show the difference. At first the original image might look decent, but the final image is dramatically better. I am totally convinced that I would not have achieved such wonderful skin tones and color balance by using the standard ad hoc method of playing with the Color Balance and Tint sliders, followed by Color and Curves adjustments. Besides, this method is much quicker and easier to use.
Figures 6 and 7 show the final numbers and the final Curve adjustments respectively. Note that the final numbers of C: 13%;
M: 43%, and Y: 44% are extremely close to the ruddy skin rule of thumb.
Let me show you two other examples. Figure 8 is an interior portrait taken with a Leica M8 under incandescent light. The original CMY readings are: C: 0%; M: 39%; Y: 36%. Using the above methodology for light Caucasian skin (Cyan approximately 1/4 of the other two values), I edited the CMY values to: C: 9%; M: 35%; Y: 37%. This change produces the much improved image in Figure 9.
Similarly, Figure 10 was shot under mixed light with a Canon G 10. When I f irst saw the image, I thought that the camera did a surprisingly good job of figuring out the correct white balance. The original values read: C: 40%; M: 68%; Y: 66%.
However, using the above methodology and editing the curves to produce the values C: 36%, M: 70%, Y 73% renders the clearly improved image in Figure 11. Note that the Cyan value is approximately 1/2 of the other two, as it should be for dark skin. This is typical; sometimes I see an image that at first seems to be very well balanced, but when I apply this methodology I always end up with a much better image. I believe you will too.
With a little bit of practice, and a few quick iterative curves adjustments, it is relatively straightforward to produce gorgeous skin tones and excellent white balance in images of people using the CMYK correction method I’ve described.