While most people do shoot RAW and can tinker their white balance settings (WB) by GUI sliders from their software (such as Lightroom, and PhotoShop), sometimes, it’s beneficial to know how to do it from scratch as well.
Of course, it’ll be best to make sure your WB is accurate to begin with, but the world is not perfect, hence image editing software are made available.
Portraits, in general, require a certain level of color accuracy to make the skin of the subjects flattering. Light skinned subjects has a tendency to look too pink with wrong WB, while dark skinned people can be too orangey.
(Click images to see a larger version)
We’ll be using CMYK (cyan, magenta, yellow, black) to reference our skin tone correction. The premise is, light skin should roughly have a ratio of approximately Magenta + Yellowvalues =5 times the Cyanvalue.
Before you panic about the equation, it’s really quite simple. If your Cyan value is 10%, your Magenta and Yellow value should add up to roughly 50%. Depending on the obvious color cast, either the Magenta or Yellow value will be higher.
It’s also worth noting that lighter skin tones will have a smaller ratio, usually a M+Y = 3x C, instead of 5x C… Conversely, dark skin tones will have a higher ratio.
These are just guideline figures as human skin tones vary greatly from race to race.
Here we go.
Here’s the original image with severe color cast from the tungsten lights in the room with the camera’s white balance at (auto) AWB setting.
1) Open your image and select your Color Sampler tool. I prefer to pick a larger sampling area (CS2 only allows up to 5×5, I think CS3 can go much higher).
Click on a non-reflective part of the face that you want to use as a reference point. I prefer the middle cheek, forehead, chin, or nose bridge in most cases. Here, I chose the area between the eyes.
I also sampled the background wall as I know that the wall is off-white with a hint of pink. If I can get that close to actual, then I know I’m close to correcting the skin as well.
2) Now look at the INFO palette, you’ll need to look at the CMYK values, so click on the little airbrush icon and choose CMYK Color from the drop-down list.
3) Our first sampling point is the skin, you’ll see the values in the red box. You’ll notice that the Y-value is really high, so we need to tackle that first.
Note the right sampling point of the wall. Since I know the wall is near-white, the number should reflect on the high-200 value.. The blue channel is obviously causing some color cast onto the wall.
So based on those alone, we know that we have to mainly fix the Yellow channel first, then adjust the others in small increments.
4) The opposite color of yellow is obviously blue. So we open a Curves adjustment layer and change the Channels to Blue (instead of RGB). Use the color dropper to click on the sampled spot and it’ll display where that sampled color is on the graph (you’ll see a square appear along the line).
Drag that line upwards (adding blue values) to reduce the yellow value. You’ll see that most of the yellow color cast is gone. The image looks more “normal” now, but we still see some pinkish cast, so we’ll need to look at the red channel next.
5) Now select the RED channel from the drop-down list and click the sampled point again to see where to adjust the curve. Once you’ve found the spot, drag the curve DOWNWARDS to reduce red value. This’ll reduce the Magenta values. Move it until you get the rough ratio above (5xC=Y+M).
Note that you don’t have to get it exact. Read the top of this page.
6) Finally, choose “RGB” from the Channels drop-down list and make a simple “S” constrast curve to complete the correction.