Getting to Know the Language of Color
by Danny Pascale, dpascale@BabelColor.com

Editor's Note
This is the second of a two-part series. Part one was published in the January 2005 issue.

In the first part of this article, we presented an overview of the characterization of color and the different languages (color notations) used for its description. This month, we will see how we can juggle between these languages. Also disccused are some of the pitfalls you must be aware of, and how you can judge the difference between two colors.

A Rainbow of Numbers
A typical question being asked is, "I would like to find the equivalent Munsell values for the #32555 FED-STD-595B color and while we are at it, I would also like to get the equivalent L*a*b* coordinates for quality control purposes, and the equivalent RGB coordinates to show on my glossy corporate bro-chure. Can you help?"

First, we have to admit that the question above is a collage of many questions, but they are all asked regularly on an individual basis. Let's start with the first part, FED-STD-595B to Munsell (a description of FED-STD-595B is shown in Table I). If you happen to have both sets of color chips in house, you can shuffle the chip cards until you are satisfied. You will not find an exact match and you will need to estimate how the FED-STD-595B color fits between two Munsell chips. The result is not instantaneous, but you will get better at it with practice. And if you have to do the reverse, find the matching FED-STD-595B chip to a Munsell color, or any other color for that matter, you will find yourself going back and forth in different sections of the FED-STD-595B book because of its peculiar way of sorting colors.

You can also use a colorimeter and measure the L*a*b* coordinates of your FED-STD-595B chip, if you happen to have one on hand. Finally, you still need to get the RGB coordinates for your printed brochure, and your graphic designer asks you for the Adobe (1998) RGB version, a specific RGB space. If you recall from the first part of this article, the RGB values for a given color are different for all RGB spaces since the various RGB spaces are built from different primaries and illuminants. This time you are not lucky, as your colorimeter does not provide RGB values.

Some instruments do, but it is often sRGB, a common all-around space, which is excellent for computer displays, but far from being able to represent all visible colors. A larger space like Adobe (1998) can represent a larger color gamut, which is still within the range of colors that can be reproduced with printing inks. After thinking about it, your graphic designer opens his graphic editing program (Photoshop, Corel Draw, Paint Shop Pro, etc.), sets its working space to Adobe RGB, and selects a color using the L*a*b* coordinates you gave him.

As you just saw, you need quite an assortment of color chips, measuring instruments, software tools, and knowledge to find the pot of gold at the end of this rainbow of numbers. Any help you can get, so much the better. Of the software tools available to help you in these tasks, most are bundled with or offered as extras to colorimeters and spectrometers. Some can only work when your measuring instrument is connected to your computer. Some can perform one or two of the conversions and matches described while others can do all and even more. The one that is best for you is a matter of the time, cost, and effort that you can assign to get a solution.

Figure 1: A screen shot of a software tool used to compare and convert colors. In this case, we see a FED-STD-595B chip (No. 32555) converted to various color notations such as Munsell HVC, L* a*b*, and Adobe RGB.

In Figure 1 we see a screen shot from one such tool, which illustrates answers to the questions presented in the beginning of this article. In the left side, the selected patch is the FED-STD-595B #32555. On the same side, near the bottom of the screen, we see its L*a*b* coordinates (74.6, 24.4, 51.0) based on the selected illuminant (D50 in this case). As well, we see its equivalent Munsell HVC notation (6.0YR 7.3/9.5). On the right side, we see that the Adobe (1998) RGB space was selected. There is also a yellow tag at the top of the display, which mentions “convert mode”; this means that the selection on the left is converted to the right settings. There is no clipping (i.e. the FED-STD-595 chip color fits within the Adobe RGB space) since the CIE94 color difference is zero (there is more information on color difference later in this article). The equivalent Adobe RGB coordinates (224, 164, 94) are numbers to be used by the graphic designer.

You may notice that the illuminant selected for the FED-STD-595B chip (D50) is not the same as the one used by Adobe RGB (D65), and the positions of the illuminants in the xy chromaticity diagram as well as the position of the colors for each side (square markers) are indeed different. However, as we discussed last month, they are perceived as the same when processed by the human brain.

Table I: FED-STD-595B The Federal Standard No. 595B (FED-STD-595B) comprises 611 colors which can be purchased as individual color chips, sets of color chips or in a fan deck. Chips are identified using a five digit numbering system ("12345"): 1st digit Finish 2nd digit Color group 1 glossy 0 Brown 2 semi-glossy 1 Red 3 flat/ lusterless 2 Orange 3 Yellow 4 Green 5 Blue 6 Gray 7 Misc. 8 Fluorescent The first digit describes the surface finish; not all colors are presented with the three finishes and many have only one. The second digit indicates an arbitrarily selected color classification grouping. The last three digits ("__345") are assigned in an approximate order of increasing luminance. The numbers are not closely packed sequential numbers; large "holes" between chips are frequent. The chips are only identified by this five digit number; no colorimetric information (i.e. XYZ or L*a*b*) is given. As per FED-STD-595B, if this standard is called as part of U.S. government procurement, it is mandatory to match a color by visual comparison with a physical chip, Therefore, you should purchase the selected chip for final approval. Chips can be purchased from a company specialized in selling standards or directly from the General Services Administration (GSA): General Services Administration Federal Supply Service Bureau Specification Section Suite 8100 470 East L'Enfant Plaza, SW Washington, DC 20407 Tel.: (202) 619-8925 Fax: (202) 619-8985

In the example we just saw, the colors are well behaved and are perceived as the same under different illuminants. But life can have darker moments. Let's say we have two objects made of different materials that have the same color under a given light. This is also applicable to the same type of material—such as plaster-wall paint—but purchased from different suppliers. Will they look the same when seen together under another light? The answer is maybe.

This can happen in many situations, such as when a manufactured object has both plastic and metal parts (think hand tools, appliances, cars, etc.). You can have a good match under natural lighting (outside), which becomes noticeably different at night under street lights. This effect, called metamerism, happens when the pigments or dyes show abrupt changes, or absorption bands, in their spectral characteristics. Because of the large coverage of each color band in the standard observer (our eyes), we tend to average these absorption bands and smooth out abrupt transitions. Then two materials that have different reflection spectrum can be perceived as of the same color under one illuminant and different under another illuminant.

From the above, you may be discouraged to use a chip card made from a material other than the one proposed in the finished product. Metamerism can be a problem only if you are not aware of it.

What is the solution? First, ask your suppliers to give you the dye and pigment contents of the product you selected. If the pigments or dyes used in the two materials are the same, there is a good chance they will look the same under any light. However, if the pigments or dyes are different, you are likely to see a difference, and you need to look at the issue more closely. Second, compare, select, and measure your colored materials under the illuminant that is likely to be used with the final product. You may need to compare them under two or more illuminants to cover many usage patterns. Third, it may be impossible, or costly, to find a perfect match for all illuminants, but at least you will have a rationale to make changes, accept the differences, and convince the marketing department that it is a design feature.

Starting on a Common Base
Standard workmanship principles should apply. In particular, for painted materials, the last coat will often be influenced by what is under it. A neutral color primer may be required if the part was previously covered by a saturated or dark color. Multiple coats, or a coat of sufficient thickness, may be needed to achieve the required color. This is on top of any issues related to surface preparation or coating thickness requirements imposed to meet corrosion or environmental criteria.

You may find all these concepts and mathematical derivations too painful and there is a solution for this problem: use a standard color system such as FED-STD-595B, Munsell, or RAL. The advantage of these systems is that you do not need elaborate measuring setups and complex instrumentation. A simple color patch will do the trick. Most are available in two finishes, matte and glossy, with FED-STD-595B presenting a few colors in a third finish, semi-gloss.

However, there are certain fundamental rules you still need to follow. The first one is to use a consistent illuminant to compare your reference and samples. A viewing booth is the best solution. As a minimum, use the same viewing conditions all the time. If possible, especially for indoor applications, agree on using the same illuminant that will be used on the finished product; this is usually well known for commercial applications.

Don’t compare colors in a room lit with fluorescents if the application calls for halogen lighting or vice versa. Select a viewing room decorated with neutral colors; a large red wall will most certainly offset your perception and give bad surprises. If there are a lot of windows, put down the shades, or go in a corner of the room where there is less interference with external lighting. On the contrary, if you use natural daylight as an illuminant, do your comparative tests around noon, and avoid direct sunlight.

Choosing the Best Standard
Is Munsell better than RAL? What about Pantone colors versus the NCS (natural color system)? Should you use an approach based on numbers only? It depends on the customer requirements, the other suppliers’ preferred methods, the field of work, the experience of everyone involved, and the available test equipment. The Munsell Book of Colors has been around for a long time and is well known around the world. For federal contracts in the U.S., the FED-STD-595B color book may be the most appropriate, even if limited in the number of available colors. When dealing with printed matter, a Pantone equivalent may be in order. For the European market, using RAL or NCS colors would be more appropriate.

Munsell equivalents in the CIE system were determined using a 2° observer (CIE 1931) with illuminant C. The latest RAL design collection is simply the L*C*h* notation presented as HLC (with h* placed as the first coordinate) but determined using a 10° observer (CIE 1964) with illuminant D65. Because of the different observer and illuminants, you cannot simply take L*C*h* values from a Munsell table and say that the equivalent RAL code is h*L*C*. A much more complex conversion is required.

If a colorimeter or spectrophotometer is available, you can specify a color as L*a*b*, or L*C*h*, or any other CIE compliant coordinates. You could then add a gloss specification, which is not limited by the available finishes of off-the-shelf chip cards.

What is the Color of Truth?
Is it better to rely on measured numbers or on hard copy chips? All color chips vendors will tell you that the only reliable way of matching colors is to use physical samples, which they sell. They will also tell you that you should not rely on number matching although most will also gladly sell you software-based catalogs with conversion tables to other color notations.

If you base your judgment on chip cards only, do you know if your card is still calibrated? It may have changed colors if exposed to the sun (ultra violet) or a reactive chemical atmosphere (as can be found in a kitchen). Do you have a chip that corresponds exactly to the color you want, or do you need a color somewhere in between two or three cards? Is the surface pristine or scuffed? Even when new, was your chip near or far from the manufacturing average for this color? Did you buy the standard quality, the higher quality, or the one coming with a calibration certificate? Many vendors do offer such varieties. How do these compare to your customer’s chip?

On the other hand, if you rely solely on measured numbers, there are also potential issues. Is the measuring instrument calibrated? What is its precision? How close are the published numbers to the colors of the average production lots? How do you take the surface finish into consideration? Finally, a numbered-based approach is not good for representing fluorescent dyes or metallic colors.

Here we have to deal with a neither a white nor black world. Numbered colors do not fade, and everybody can easily share the same reference. They are ideal to make rapid computer-assisted match selections and decisions. You don’t need to buy multiple collections of color chips. Once you have targeted a few good candidates, you can order the chip cards in larger sizes to make your final assessment and control. Certain standards, such as the FED-STD-595B, make it mandatory to make the final judgment on a physical sample, and you may be bound by it if contractually required.

Should you rely on the colors seen on a computer screen? In most instances, no, unless your display is calibrated. There are many devices and software tools—an entire industry is dedicated to this task—and prices for calibration systems start in the few hundred dollars range (some professionals also offer calibration services). While this industry is mostly concerned with the printed world, a calibrated color is a calibrated color, whatever its field of use, and the accuracies required in the print field can be quite demanding. However, the advice relative to specific surface finishes and to the other caveats mentioned before still applies. In many instances, a match made on numbers only is sufficient. It all depends on the required matching tolerance and how critical the other appearance features are.

Now, if you agree to the principle that the delivered color will be different from the required one, you need to define an acceptable margin. Color differences are expressed as DeltaE, a value based on the Euclidian distance (the shortest line in 3D) between the coordinates of the reference and sample (the word “Delta” is often shown as its Greek symbol (D), a small triangle; see last month’s article for an illustration of how DE is determined.) There again, multiple standards were developed, and newer ones are being looked at.
The older ones, DeltaE*ab (DE*ab) and DeltaE*uv (DE*uv) are directly associated to the coordinates of the L*a*b* and L*u*v* representations. Efforts in making this color difference even more uniform have brought us the CIE94 and CMC color difference formulas, both derived from L*a*b* data. More recently, the DeltaE2000 color difference has generated increased interest.
In all of these, the goal is that a DE = 1 corresponds to a barely noticeable difference when viewed by 50% of the population. That is, the difference formula should work equally well for whitish-blue colors, saturated reds, and dark brown ones, a very difficult task indeed. CIE94 is often considered a “better” choice for small color differences than DE*ab, which is a “good” all around choice, especially for large color differences. Be aware that CIE94 values for small color differences are about half the ones obtained with DE*ab, so specifying which color difference formula to use for comparison is important.
While a color difference of one opens a room for debates, a color difference of five will be noticed a lot more rapidly by more people. A DE in the range 1 to 5 is a good place to start. More critical graphic design or paint touch-up applications may go toward the low values, whereas applications where it is difficult to compare the color with a reference could go over five.

Figure 2: Comparison of FED-STD-59B

Figure 2 compares the FED-STD-595B chip #32555 with its closest Munsell match 5YR 7/10. The CIE94 color difference is 3.23 with the difference coming from a mix of lightness (DL* = -2.45; i.e. the Munsell chip is darker), Chroma (DC* = 3.62; i.e. the Munsell chip is more saturated), and hue shift (Dh*= -3.33 degrees; i.e. the Munsell chip is redder). The actual L*, C*, and h* values of each chip can be seen on each side of the figure. The chips are also reproduced in the two concentric square patches, the larger square corresponding to the left side and the smaller one in the center corresponding to the right side.

Understanding the Nuances
We have seen that there are many ways to express a color specification. It is important to understand the concepts of observer and illuminants and to clarify which color difference formula will be used when dealing with numbered values. All involved should agree on achievable accuracy goals and on the method with which they will be determined. The selected method, either measured or visually compared, or both, can vary from the simple to the very complex and should be dealt with according to the criticality of the color specification.