One of the most powerful Photoshop tools at your disposal is Curves. Though it’s often only used to tweak contrast, the curves tool is also hugely effective in correcting color. What’s more, learning how to use it gives you a greater knowledge of image editing in general. It’ll enhance your understanding of the histogram and teach you how to edit photos by numbers.
The idea of correct color
Correcting color in images is about removing unwanted color casts. The “unwanted” part is important because some color casts are desirable. For instance, you wouldn’t want to neutralize the warm hue of a sunset. However, you might want to remove the blue color cast that sometimes pervades photos taken on overcast days or in hazy conditions. By removing an unwanted color cast, you’ll reveal the true color of the objects and subjects in your photo and make the image “pop”.
The content of a photo will dictate how you edit it, so you shouldn’t obsess over correcting color in every photo. Many times, you’ll want to do little or nothing to the color. An appreciation of the curves tool and the numbers around it will help you decide what each photo needs.
A neutral histogram
When working with curves, histograms, and RGB numbers, it’s useful to know what the histogram is telling you. There’s no such thing as a right or wrong histogram, per se, since it only mirrors the pixel data of the image, but it will highlight potential problems.
By looking at all three RGB (red, green and blue) histograms at once, you can immediately get an idea of whether or not the image has a color cast. If there’s no color cast, the three histograms will look very similar. A black and white RGB image illustrates this perfectly because it’s completely neutral. In that case, the three RGB values will be equal in every part of the image and the histograms are identical.
In this black and white RGB image, you can see that the red, green and blue histograms are identical. You won’t get this in a color photo, but if the histograms look similar, especially from the middle to the right-hand side, there is unlikely to be a noticeable color cast.
Getting ready
Before you start in curves, there are a couple of things you’ll need to prepare in Photoshop:
- Be sure that the “Layers” and “Info” windows are open.
- Select a “3 by 3 Average” or “5 by 5 Average” sample size for the eyedropper tool.
Easy one-click color correction using mid-tones
Whenever a photo contains an area that should be neutral gray in your estimation, you can use the mid-tone eyedropper tool in either levels or curves to quickly correct any color cast. Simply clicking on the supposedly gray portion of the image will correct the color. It’s usually worth clicking a few times in different areas until you achieve a result that pleases you. There are ways of calculating precise mid-tones in an image to make this method more precise, but guessing often works well and is far quicker.
By dragging the eyedropper tool over this gravestone, we can see that the green value is less than that of red or blue. We might reasonably expect this stone to have a neutral gray color, which would give roughly equal RGB values, but the green deficit (RGB 160, 149, 160) indicates a magenta (opposite of green) cast.
By opening curves or levels and clicking on the stone using the middle eyedropper, the magenta cast is corrected. As a result, the green in the photo is stronger.
Using curves and the info palette to correct color
By introducing the info palette into the equation, you can make far more precise color corrections. The technique you’re about to learn also teaches you to evaluate and edit photos by the numbers. Think about this – when you have nothing to compare an image to—no alternative version—it often looks “okay” at first glance. By studying the RGB values, you’ll get a clear idea of any potential problems in the photo.
Before you proceed, it’s important to note that an image always needs “neutral” areas for color correction to succeed. That’s because a neutral tone provides a known reference point that you can work from. Neutral pixels always have identical RGB values (e.g. 128, 128, 128). Any photo that doesn’t contain a neutral tone is difficult to accurately correct. This is true whether you’re adjusting color yourself or hitting an auto-color-correct button. Photographers often use gray cards to introduce a known neutral into the image for color correction later.
10 Steps to Color Correction with Curves
Here are the steps you might take to correct color using curves, the info panel, and histograms:
Step #1 – Select the eyedropper
With your image open, select the eyedropper tool from the Photoshop tools palette.
This is a picture of a hotel in Switzerland before any color correction. There is no strong color cast, but you might detect its cool bias.
Step #2 – Check white RGB values
Hover the eyedropper tool over a diffuse white highlight in the photo with RGB values in the 230s or 240s (try to avoid high 250 values). Use the info palette to see these values.
Hovering the eyedropper tool over diffuse white highlights, I can see the blue channel has consistently higher numbers than red or green. The difference isn’t drastic, but it does indicate a blue color cast.
Step #3 – Create a sample point
Hold down the Shift key and click to create a sample point from this white area, which will show in the info palette as #1. It’s possible to move a sample point after you’ve created it by holding down the Shift key and dragging.
Step #4 – Repeat with mid-tones
Repeat this procedure with a neutral gray mid-tone, if you can find one, with RGB values of around 120-140.
Here, we’re creating a mid-tone sample point. Again, you can see from the numbers (RGB 111, 120, 137) that there’s a strong blue presence. The highlight sample point that I’ve only just recorded is stored on the depicted info palette at the left, third down (marked as #1).
Step #5 – Repeat with shadows
Do the same thing with any black, shadow areas with values of about 10-30. After that, you’ll have created three sample points. Since color casts in shadows are inherently harder to see, this third sample point can often be skipped without ill effect.
A shadow sample point from the trash bag records at RGB 20, 21, 29. We can see now that a cold color cast runs through the entire image from highlights to shadows. Once I’ve clicked on this, my three sample points are all stored and displayed in the info palette.
Step #6 – Analyze the three samples
Looking at the three RGB samples you’ve created, you should get an idea of any color casts that are present. You’ll typically see the same problem across all tones from highlights to shadows, though not always. Remember that a low RGB value in any of the three channels indicates an opposite color cast. Thus, a low red value indicates a cyan cast, low green is magenta, and low blue is yellow. This only applies in areas that should be neutral in color (i.e. white, gray, black).
Step #7 – Open a curves adjustment layer
Open a curves adjustment layer. Hold down the Ctrl and Shift (Cmd + Shift) keys and click once again exactly on the center of the second, mid-tone sample point you created (#2). This has the effect of placing a mid-tone point along each of the individual RGB curves.
Holding down Ctrl + Shift (Cmd + Shift in Mac) and clicking with the eyedropper tool places a sample point on each of the three RGB curves. This is useful for adjusting specific mid-tones. Here, I’ve opened the red channel to illustrate this. To correct color, you have to adjust the individual red, green and blue channels until the corresponding output numbers on the info palette match.
Step #8 – Correct the color cast
Now it’s time to correct the color cast. On a curves graph, the top right point represents highlights and the lower left shadows. In between are any mid-tone points that you placed on the curve.
Starting with highlights (your #1 sample), open the individual red, green and blue curves channels one at a time and move the top right point either left or down along the outer edge of the graph so that, eventually, the three values match. As you move each point on the graph, the info palette gives you the updated output value.
Usually, it’s best to choose the lowest or middle of the three existing highlight values and match the other two to that (see “tip” below).
This is an exaggerated example of a curves highlight adjustment in the blue channel. I’ve pulled the highlight point to the left, which has added blue to the image (far too much blue). Moving the point down the right-hand side of the graph would increase yellow. In reality, these edits will usually be very slight, moving only a small amount either way. The info palette will reflect these changes in the RGB output numbers.
Step #9 – Repeat for all three points
Repeat this process with the mid-tone and shadow points, so that all of the chosen neutral points in the image are in fact neutral. The bottom-left shadow point is also moved along the outer edge of the graph, either upwards or right. The mid-tone point you’ll drag either up or down. If the color looks wayward at the end of this process, it typically means that you’ve picked a sample point that wasn’t neutral. Ensure that your sample points contain no color noise or reflected color. Zoom in on the area you sample to make certain of this.
Here you’ll see that all three sets of RGB values have been equalized – for each point, you see the before (left) and after the (right) value (Red on point #1 went from 238 to 239, Blue from 244 to 239).
Now all points are roughly similar, in other words, all the sample points I took that were estimated to be neutral have been made neutral. Note that the numbers don’t have to match perfectly like this as long as they’re close. In the curves graph, the red channel has been lifted and the blue channel pulled down slightly as a result of my edits. The green channel was untouched in this instance, so the corresponding line cuts straight through the middle.
Step #10 – Remove samples and save
Once the correction is complete, the sample points are removable by holding down Ctrl + Alt (Cmd + Option) keys and clicking on them. You should see the scissor icon when you hold these keys down. To finish, either save the image with its adjustment layer intact or flatten the layers, as required.
Before – uncorrected image.
This is the color-corrected image. With the blue cast gone, other colors in the photo can breathe. In particular, the yellow color of the hotel and the green in the grass and trees are more prominent.
Tip: Since moving the endpoints of the curve line affects all highlights and shadows, you should edit conservatively. In particular, avoid choosing the highest of the three RGB values as the target when matching red, green and blue highlight channels. Otherwise, you may find that you blow out wanted detail in the brightest part of the image. Flaws in the shadows are generally less noticeable, but you still risk blocking detail if you adjust all the shadow RGB points to the lowest of the three values. In general, turn the numbers away from their extremes.
Mixed Lighting
The types of correction discussed in this article work best when there are naturally occurring color casts in the image. In mixed lighting, where the light sources are radically different (e.g. incandescent lighting and daylight), you’ll need to painstakingly address each affected area of the image using layers in Photoshop or the adjustment brush in Lightroom. Avoid this type of lighting wherever possible, since it’s difficult and time-consuming to correct in processing.
Finally
I don’t expect that you’ll use these techniques on every image, but I hope they’ll improve some of your pictures and that you’ll enjoy experimenting using curves in Photoshop. This type of mathematical editing gives you a good understanding of histograms and the meaning of RGB values.
Merely hovering the eyedropper tool over a picture while watching the numbers will tell you something about it. If there are no naturally occurring “neutrals” in the photo and you want consistent or accurate color, a high-quality gray card provides a solution.
Please don’t hesitate to fire questions my way if anything is unclear.
