Using Curves to Separate Grayscale Images

The Curves menu in Adobe Photoshop is a favorite of mine when it comes to separating a very detailed illustration for screen printing, because it allows me to extract the information in a systematic and structured way while minimizing the loss of tonal range. Many of the other methods of pulling colors from an image in Photoshop tend to be choppy, inconsistent, and even ineffective depending on an image’s color composition and tonal values.

The Curves menu in Adobe Photoshop is a favorite of mine when it comes to separating a very detailed illustration for screen printing, because it allows me to extract the information in a systematic and structured way while minimizing the loss of tonal range. Many of the other methods of pulling colors from an image in Photoshop tend to be choppy, inconsistent, and even ineffective depending on an image’s color composition and tonal values.

The Color Range command and the Magic Wand tool are among the most popular methods of separating color selections in Photoshop. But both of these approaches fall short with images that are highly blended, use overlapping transparencies, or contain colors with similar values that are close to neutral grays. However, these tools have their place in color separation. The Color Range command is a very useful method of extracting complex selections from images, and I commonly use it first to see what I’ll get, even when it’s not the technique I ultimately decide to use when creating a selection for a color channel.

The Color Range tool will often create very choppy selections of colors when you have an image with a lot of gradients blending into each other or into the background (shadowy effects). The Magic Wand tool is really best applied when the image has large areas of flat colors that you can easily select and isolate. Any blending or variation in color tone tends to throw this tool out of whack so that it either selects not enough information, or everything around the color that you need.

The Curves menu works better with very detailed designs than these other options because the Curves menu always starts with all of the information available in a specific area that has the curve applied to it—not just a portion of the information inside a selection. In essence, this is always preferable to a second-generation copy of the information, which is slightly worse than the original. To use the Curves menu properly in the separation of colors from a design, isolate areas for preparation, pull practice selections for testing, and then create a final set of curves to extract the necessary colors.

Isolating areas for preparation

The more colors in the design, and the more complicated it is, the more areas you’ll have to isolate for the preparation stage of separating. You should save a complicated design with every major area in it as a separate layer whenever possible. This method can simplify the use of the Curves menu in isolating and separating. Neglecting to save a complicated design can lead to many hours of work to correctly select specific areas for color extraction. For this reason, the curves method is sometimes a less efficient way to separate artwork. The most common issue at the beginning stages of isolating a design for separation is selecting image areas. Each design is different, but a good way to approach this issue is to first address the theory behind the curves method and the supporting concept of tonal range.

The Curves menu depicts tonal range with the absolute black point on one corner and the brightest white point on the opposite. The function of an effective tonal range is to demonstrate a noticeable division between the values on a typical grayscale. Many screen shops fight to save the tonal range from dot gain, moiré, screen-washout issues, and printing inconsistencies. What commonly occurs from tonal-range problems is tonal compression or the loss of the ability to differentiate between values in the highlights, midtones, or shadows.

When you look at a basic gray-scale generated on the computer next to a grayscale that is damaged by tonal compression, you will note the loss in the midtones and highlights first (Figure 1). You’ll have trouble discerning where the 50-60% line is because the midtones and highlights tend to merge together. To avoid this issue, use a pure grayscale depiction of a specific color value that you can then isolate to produce a superior final separation set. The idea is to first isolate a selection of color and then create an ideal grayscale from it that can then be depicted in the separation set.

A simple way to start to understand the curve separation process is to use a grayscale image and pull several colors from it. An illustration that I made for a local martial arts club fundraiser is a good example of a design that can be separated by using the Curves menu to recreate the subtle tonal range in a grayscale image (Figure 2). You won’t have to isolate a lot of an image that is already in grayscale, particularly when the image will be printed on a black background. I didn’t need to isolate any colors in this example.

Pulling practice selections for testing

Doing extra work for reference selections paid big dividends when using the Curves menu to separate my design. Creating a separation from the Curves menu can be very difficult—if not im-possible—without some kind of ballpark reference selection to guide me in getting the right start and finish points.

In the example presented here, I decided to use a common solution for detailed grayscale images on black shirts: pull three gray colors (dark gray, medium gray, and light gray) and a highlight white. I used the Color Range tool to quickly extract selections that I saved as alpha channels for each of the gray colors that I intended to use. I created an additional alpha channel, moved it in front of all of the other channels, and labeled it “shirt” with the Channel Color option selected as black. Next, I opened the Color Range menu for each gray channel and used an appropriate gray value to select that section of color from the original design (Figure 3). I really had to watch the Fuzziness selector as I slid it up and down on the Color Range tool so the gray selections I created would overlap each other slightly. If the selections hadn’t flowed smoothly into one another, the final design would have been clunky. Being picky with a reference channel can prevent the recreation of mistakes in the final curved version.

I saved each selection into an appropriately labeled alpha channel with the proper colors selected in the Channel Options menu so I could view how the selections would look against the shirt background and each other when I made the channels visible. In the end, I had a quick set of separations that I could use as a reference when using the Curves menu to create my final set of separations.

Creating the final set of curves

The tricky part of separating this image by using the Curves menu was finding just the right curve that would isolate the selection of color from the file. I left the reference file open for the light-gray separation and duplicated the original file. Next, I opened the Curves menu with the duplicate file open (Figure 4), slid it off to the side so I could view the color-range selection underneath, and slowly created a curve that isolated a selection similar to my color-range selection. I used some general rules for forcing the Curves menu to do this kind of extraction:

• Make a smooth curve whenever possible. This seems to create the best separation file.

• Moving the black point all the way to the top (or bottom if your Curves menu is set up the other way) to knock out the black.

• Find the point on the curve where the gray level is close to the value of the selection for extraction. If you want to extract a 50% gray, then take the center of the curve (the 50% point) and move it entirely away from the knocked out black point. This creates a new 100% black point, meaning that my 50% gray is now black and my black has now become white by changing the position of these appropriate points on the Curves menu. This is similar to inverting the file, which you can do by shifting the white point all the way up and the black point all of the way down, except that this process inverts the file at a gray point rather than a white.

When the final curved files looked good, I copied them and pasted them into duplicate channels right next to the original test files that I pulled using the Color Range command. I could then view the channels in proper print order and check them against the test files that I used for reference.

Additional curves were necessary, and I still needed to create the highlight white. I applied additional curves to the pasted channels in the midtone regions to adjust for dot gain. I just bumped the midtone point of the curve up 10% to compensate for gain and saved. I quickly created the highlight white by duplicating the original, pasting it as a channel, inverting it, and then sliding the white point over and curving it down to knock out most of the midtones while still leaving a smooth transition to the other colors (Figure 5).

I’m always surprised by the amount of extra subtlety that I can create in the files by using the Curves tool instead of Color Range selector. I believe it’s easier in the long run to do the extra work on detailed designs and separate using the Curves menu because other methods, such as color range, calculating channels, or converting the image to quadtone, cause the separation set to lose information. You can’t retrieve the information once it’s lost. For this reason, I suggest you investigate the use of the Curves menu when separating complicated images that require the controlled reproduction of fine detail.