Describing, Specifying, and Using Digital Digital Color Space
Discuss the concept of digital color space and how the color models used in today's graphic-design software can help you manage color at each stage of the printing process.
Until recently, accurate color reproduction for screen printing and other graphic-arts fields relied solely on skilled press operators and other technicians on the production floor, who adjusted the printing process for various substrates, inks, and applications to achieve the colors specified by customers. But the range of control these individuals had in manipulating color was limited by the adjustment range of screenmaking tools, presses, and other elements of the process. It was a highly subjective approach to color reproduction, and it remained largely unchanged over the last 150 years.
With the advent of digital desktop color, however, the ability to specify, calculate, and manipulate color has moved to the front end of the reproduction process--the graphic design stage--and reached new levels of control. Presses and other traditional analog tools no longer command the weight and consideration they once did.
The implications for the screen-printing industry are enormous. In today's market, designers are working with artistic and commercial elements that may be used for multiple applications including fleet marking, P-O-P and retail display, textiles, trade-show exhibits, decals, and signage, not to mention non-screen-printed applications like catalogs, brochures, and even the World Wide Web. Digital graphics tools make it possible to provide accurate, consistent color across this range of materials and visual formats, a critical requirement in today's competitive graphic-arts environment.
The need for color management
Why do screen printers need to understand and apply color management? The reason is quite simple: Advertisers want their images to be consistent, and color management assures this consistency. Screen printers compete against other's processes, including offset and digital- printing technologies, that are already capable of working within managed color environments. So if a designer can specify that an image be produced with screen printing or a wide-format digital printer (e.g., inkjet, electrostatic), and only the digital-imaging system is color managed, it's likely that the digital system will win the work.
The adaptation of managed color has fallen mostly under the framework of the International Color Consortium (ICC), a body made up of numerous color professionals from all areas of the graphic arts, including photography, instrumentation, scanning, computer technology, printing, display technology, software, and other industries. Although the process of developing coherent color-management strategies is still very much in the formative stage, hardy early adopters are already making it work in daily production. But the complexities of color specification, characterization, profiling, and transformation are still a big reach for the average screen-printing plant.
This article is designed to introduce you to key aspects of digital color and how various color models are used to control color reproduction. It describes each of the color-specification models and explains appropriate uses for each. Additionally, it defines common color-management terminology and addresses the issues involved when translating between color models.
Characterization and profiling
At the core of the color-management issue is the concept of device-independent color. This means that color speci-fication is not limited to any one device. A device might be a digital camera or a scanner for capturing color, a monitor for viewing color, or an output device (or process like screen printing) for reproducing color. Since you are moving a digital color image between a number of these devices, you need some way of preserving the continuity of the color at each step. The solution is using digital technology to establish and maintain independent color space.
The processes by which you apply digital color management to specific devices are called "characterization" and "profiling." While this article won't cover the mechanics of these processes, it is important to understand what they do. Characterization tells us what level of color representation each particular device or step of the process can achieve, and profiling brings each device or step into an acceptable range for image capture, viewing, or output.
The profile for each step is used by the design software (e.g., Adobe Photoshop 5.02, QuarkXPress 4.04, and others) to assure that the final reproduction of the image will be predictable. Note that this predictable output is not necessarily the optimum reproduction. But it is the best possible color reproduction based on the maximum range of color (or gamut) supported at each step of the reproduction process. The profile locks down this gamut for every step of the process.
Color can be described in many ways, and most of them have their roots in the artistic blending of color. Remember Art 101 and the traditional color wheel you studied there, which showed how primary, pigmented colors could be mixed to create additional colors? This approach was quite satisfactory in analog work, but is woefully deficient in today's digital mode.
To begin with, the cones in the human eye are sensitive to red, green, and blue light, and you perceive light when those cones are stimulated. The brain blends and interprets these stimuli as a specific color. RGB values mean little when it comes to describing color because these values aren't directly tied to any particular color.
Human vision has inherent limitations, and one of the biggest is that your eyes are "adaptive." This means that what you see is not an objective record of the scene. Your brain adjusts what you see to represent what it should look like. If you need evidence, consider what happens when you take a picture of your child blowing out the candles on a birthday cake in a dimly lit room. The scene looks perfectly natural to you when you experience it, but when you get the photos back, you notice a distinct orange color cast over the entire picture.
When you viewed the scene originally, your mind had adjusted the "white point" of the candle light to represent a "natural white," and adjusted your perception of all other colors in the scene accordingly. In reality, however, candle light is quite orange, and because the emulsion on the photographic film is sensitive to the spectrum of visible light, the photo is a much more accurate representation of the scene.
All of the color models you'll learn about here have some deficiency. Each is designed to solve one or more problems in calculating color, helping you interpret and predict the way viewers will perceive the final graphic. This is important to recognize because it reconfirms that color perception is subjective.
No matter how careful you are, it simply is not possible to make color management a "push-button" process. It will always require some degree of interpretation. Any given color model is only accurate relative to the specific color characteristics the model was designed to simulate. As soon as you begin to think in terms of absolute color, big problems develop.
The goal is to find a better, more logical way to position color, a way that makes sense. One way is to describe color in terms of three primary components:
1. hue, which is the name of the color
2. saturation, which is the purity of the color
3. lightness, brightness, or value, which represents how bright or dark the color is
This approach results in a family of useful color models that are easy to visualize, but not particularly accurate. The artistic description of color has yielded a more scientific approach.
Over the last 70 years, a body of scientists, known as the Commission Internationale de l'Eclairage (CIE), has worked diligently to quantify how we see color and derive mathematical models by which we can measure, calculate, and represent color. CIE's work is the foundation on which modern color computation is built. All of the color models presented here are derived from traceable CIE XYZ values.
The first work done by the CIE was to characterize the "standard observer." This was done by averaging the visual responses of 2000 non-color-blind people to controlled viewing of color samples. From this data, CIE developed cone sensitivity profiles for different wavelengths of red, green, and blue light, known as RGB tristimulus values. From this, the group derived the XYZ color model, described by a graph that plots the relative sensitivity of each of the three cone types to the different wavelengths of light (Figure 1).
On paper, this two-dimensional diagram represents an esoteric chart of what we actually see. If we describe a color by its X,Y, and Z values, the resulting numbers mean little more to us than three values for the relative amounts of red, green, and blue light that make up the color. To a color scientist, however, these values form the basis of a broad mathematical process for describing color accurately.
While the RGB tristimulus values describe the component mix of color, graphic-arts processes require a more intuitive model that is easier to visualize. Color representation varies tremendously between models; some color models (like XYZ) cannot be visualized, while others can. But no single color model has all of the qualities that would make an ideal representation.
Part of the problem is that color models based on human perception are nonuniform. The relative distance between colors when depicted in these models is inconsistent. This is due to the fact that the sensitivity of the human eye is not uniform. Color models based on the eye characteristically depict a higher sensitivity to light colors than they do for dark, and higher sensitivity to green than other colors. This makes it difficult to translate color from this model to a more uniform or easily visualized model.
Conversely, uniform color models do not represent the relative distances between colors as the eye sees them. This leads to gamut "clipping," meaning that the range of color represented is compressed and that there will be a definite loss of visible color. In other words, colors that you can easily see are not represented in the color space.
But this is not necessarily bad. If you chose a color model that describes the color range of printing inks, dyes, or toners, the gamut of available color is already far less than what you can see. The simple truth is that no pigments or dyes exist that can be used to make all visible colors. Your goal is just to match the color model as closely as possible with what you can actually see.
Key color concepts
When it comes to color modeling, you need to consider several points:
Color families Colors are grouped in two families: additive color and subtractive color. Additive color works on the principle that the addition of red green blue light, in the right proportions, will cause your brain to sense white, while different proportions lead to other colors. The absence of these three colors creates black.
Subtractive color works with reflected light. It represents how all pigments and dyes work. When white light illuminates a surface, the reflected color represents what is left after the pigment in the surface absorbs or subtracts certain wavelengths from the light. This is the principle upon which CMY is based. A surface that appears green to us is really absorbing magenta (blue red wavelengths) and reflecting green. In this family, black is created by a surface that completely absorbs all of the wavelengths of white light.
Color model types Color models fall into one of three categories. The first includes emissive models, used when an energy source is emitting wavelengths of light, which you sense as color. A computer monitor is a good example of an emissive device.
Models in the second category, transmissive, are used when light is transmitted through an object to achieve a particular color. These models apply to items such as transparencies, slides, backlit displays, and stained glass windows.
The third model type is reflective, and it is based on surface absorption of light. What you see are reflected components of the white light. Most screen prints (excluding backlit images) fall into this category.
Color gamut This is the last major concept involved in color modeling. It refers to the range of perceived color for any given color model. The size of the gamut is important. You can always reduce the gamut of colors in an image, but you cannot increase the gamut once it has been compressed. This is important because every type of reproduction has a smaller gamut than what the human eye perceives.
When you plan to reproduce a colored image, you need to know the size of the gamut for any given color space or any given device. To have entirely consistent color across many different devices and models, you must plan around the device or production step that has the smallest gamut. That way, you're assured that the gamut will be supported by all the larger-gamut elements of the process.
Commonly used color models
All the following color-space models are CIE derivatives. As such, you can translate color back and forth between them. However, these translations may result in data or color loss. The gamut size and uniformity of the color space will dictate how much color information is lost. The more nonuniform the shape of the color gamut, the greater the likelihood of color loss or gamut clipping during the reproduction process.
XYZ This is the granddaddy of all models and directly traceable to the RGB tristimulus values of the standard observer. It is a device-independent, nonintuitive space that is difficult to visualize. In this model, X corresponds to the red curve, Y to the green, and Z to the blue.
The XYZ model most closely approximates human vision by mixing RGB values to create additional colors. It is capable of representing hue, saturation, and lightness or brightness using the three values. As shown in Figure 1, XYZ values can be plotted on a graph.
Colors measured with a spectrophotometer are commonly expressed in terms of the relative intensity of their XYZ components. Saturation is represented by single or double peaks of high intensity wavelengths. Desaturated colors are represented by relative uniformity across all wavelengths. Brightness is represented by high peaks across the spectrum. Dark colors are represented by low intensity values across the spectrum. If the intensity values are high, the color will be heading toward white. If the values are uniformly low, the color will be heading toward black. Spectral values for three sample color are shown in Figure 2.
xyY This model is derived directly from XYZ, but the Y now represents value or luminance. The x and y represent coordinates in the resulting chromaticity diagram (Figure 3). They are derived as: x = X/(X Y Z) and y = Y/(X Y Z).
The xyY model is most commonly used to present a closed plot of the chro-maticities of human vision. Purple (ma-genta) is not a naturally occurring color and is made up of both red and blue wavelengths. This xyY plot joins the end points of red and blue to form the familiar horseshoe shaped plot. This is another nonintuitive space that is used heavily for plotting the positions of colors relative to other colors.
One of the most common uses of the xyY model in the graphic arts is in the Printing Ink Set-Up segment of Adobe Photoshop. After measuring singular CMY values, the overprints of C M, C Y, M Y, and C M Y, the xyY values for these inks are entered into the custom table, which is then used to build separation tables in Photoshop for that particular series of CMYK inks for the particular substrate on which they were measured.
The xyY values are also often used to specify the phosphor chromaticities of the RGB electron guns in your moni-tor. These values are used in conjunction with monitor calibration software to profile your monitor.
L*a*b* The L*a*b* color space is one of the most widely used and useful of the color models. It is a device-independent space as well. It is the main calculating engine in Photoshop and is used as the basis for all mode transforms. It was developed by the CIE in 1976 as a uniformly perceptual color space. It has a very wide gamut, comprising some 2,381,000 colors. While smaller than the RGB gamut, L*a*b* effectively envelops the gamut of almost all printing processes.
The L value represents luminance, and is perceptually modeled after the roughly logarithmic characteristic of how the human eye senses value. The L value ranges from 0 for black to 100 for white in uniform steps.
The a and b values are represented as a, -a, b, and -b. The axis of color for a is red/green and for b is blue/yellow. The values range from 127 to -127 for a and b. This is a difficult color space to visualize. The numbers are difficult to reference without a chart.
The important point is that L*a*b* is very uniform and calculations made within it are accurate and require minimal computational overhead. L*a*b* is used extensively in coatings and appearance measurement. It is probably the most commonly used color space for industrial color measurement and control.
In recent years, an effort has been made to scan directly into L*a*b* color space for the acquisition of digital images. This will become increasingly important as more and more color separation work is done in "n Channel," where the separation ink colors are not linked to the traditional RGB or CMY colors used in the past. Of all the color space models, this is probably the one you should learn about because of its already wide acceptance and application.
LCH The LCH model is a derivative of L*a*b*. The L represents luminance, just as in L*a*b*, with a range of 0-100, black to white. The C stands for chroma and is representative of saturation or purity of color. In this scale, the chroma range is 0-141. The H stands for hue and is measured as hue angle.
The L and C components are readily understandable. The hue angle is a new twist. This works on the principle of the color wheel. The origin is 0