Removing a Color Cast

So far we've worked only on the composite RGB channel, but you will remember from Chapter 2, "Channels," that you can also work on specific color channels. This ability to target specific zones of specific channels is the foundation for color correctionif you have a red color cast, you pull down the curve of the Red channel.

Before we send out a posse to round up all our images with color casts, it's worth reflecting that not all color casts are bad. An orange color cast in a dramatic sunset picture, a yellow cast from late afternoon "magic light," a red cast from a concert photo taken under stage lightsthey're all good. Bad color casts are the green complexions of people photographed under foliage, the jaundiced look of shots taken indoors under electric light, or the blue-rinse look your photos may have if the white balance of your camera was incorrectly set.

There are two basic approaches to removing a color cast: by the numbers and subjectively.

Color Correction by the Numbers

When color correcting by the numbers, you identify an area of your imagepreferably a midtone areathat should be neutral gray. Neutral areas are made up of equal parts red, green, and blue. It doesn't matter exactly what the numbers are, just that they are the same. Adjusting individual color curves makes it possible to equalize the RGB levels at any point on your image, thus neutralizing these areas. Making such a change in the midtone area is usually enough to fix the color cast throughout the image, but if necessary you can also adjust highlights and shadows by shaping your curve accordingly.

The easiest types of images to correct by the numbers are those that include reference colorsareas we can confidently say should be neutral. Concrete or tarmac, for example, should be gray, and gray is neutral. Using your Color Sampler tool you can sample such an area. Then, using a Curves adjustment layer and your sample point and Info palette as references, you can make the RGB numbers equal to neutralize the color cast. Let's look at an example.

Figure 5.9. The original image has a strong blue cast (example A. In the color corrected version the Red, Green, and Blue values are equalized in the midtones (example B).

1.	Open the Wine Cask image.
2.	Using the Color Sampler tool, click the foregroundthe concrete floor in front of the barrel. This will sample the color values at this point and display those values in your Info palette. To get the most accurate sample, set the options for the eyedropper to 3x3 Average.
3.	Click the New Adjustment Layer icon at the bottom of the Layers palette and choose Curves. My sample shows that the Red and Green are similar in value, with the Green slightly higher. The Blue is much higher than both the Red and the Green. I want to bring the Red up to the Green, the median value, and the Blue down to the Green.
4.	While viewing the RGB curve, lock a point on each of the color curves by holding down Command/Ctrl-Option/Alt and clicking again on your sample point. Locking a point takes the guesswork out of finding exactly where the sample point occurs on each of the color curves. Because changing the Blue curve will have the most dramatic effect, let's do that one first.
5.	To neutralize the Blue channel, choose Blue from the pop-up menu at the top of the Curves dialog box. Click the point on the curve; the Input and Output values are currently the same. The Input value should correspond to the Blue value of the sample point on your Info palette. If it doesn'tand it's possible you may be a digit or two off depending upon exactly where you clicked in the imagechange the input value to match. Make the output value of the Blue curve the same as the input value of the Green curve. This change should go most of the way towards fixing the color cast. Figure 5.10. Reducing the Blue curve (example A) and the Red curve (example B).
6.	Repeat step 5 for the Red curve, changing its Output value at the sampled point to the Input value of Green. The Output values for your Red, Green, and Blue should now be the same (or within 1 or 2 levels of each other).

Subjective Color Correction

Sometimes a by-the-numbers approach doesn't cut it, and you'll want to tweak the color curves subjectively for a result that matches your expectations of the image. For example, in this desert shot, I know I want the sand to be more yellow. To add yellow I reduce its color complement, blue, by pulling down the Blue curve.

Figure 5.11. The original image (example A) and the corrected version (example B). The Blue curve is pulled down to make the sand more yellow (example C), and the RGB curve is tweaked for more contrast (example D).

Color correction is a complex and fascinating topic that I don't have the space (or the experience) to get into in any detail. But here's a useful quick guide. To remove a color cast, either reduce that color curve or add its complement

Figure 5.12. Color complements are at opposite points of this six-pointed star.

For example, if you're working on an RGB image that has too much red, simply pull down the Red curve. If the same image were in the CMYK color mode, which has no Red curve, instead increase the Cyan curve. If an RGB image has a yellow cast, pull up the Blue curve to add blue, so reducing the yellow; if the same image were in CMYK, simply pull down the Yellow curve.