Daniel L. Randall, Datacolor, Lawrenceville, NJ

Introduction

Since the introduction of spectral-based imaging systems some five years ago, the communication of lab and production submits between retailers and suppliers has proven to be one of the primary economic applications of the technology. Recent advances in color curve generation and image processing provide opportunities for additional improvements in areas of collaborative color development, color marketing, and color prediction in multi-step processes.

Current State of Imaging Technology and Application

The development of spectral-based imaging as a technology began in the mid 1990’s, originally as a consorted effort at UMIST (U.K) for the purpose of achieving precise on-screen color of textiles. The consortium, composed of Marks and Spencer’s, and a number of it’s suppliers, were interested in the display and printing of color on texture to an accuracy and precision not previously achievable using conventional color profiles. Much of the work was focused on the communication aspects of color to reduce time to market. This is by far the application with the most economic benefit to retailers who wish to approve colors for lab dips and production without having to wait on shipments to arrive. There are currently a sizable number of retailers and branded apparel companies utilizing these imaging systems to communicate color via email with their suppliers. Many of these firms have reported their results in a recent AATCC symposium1.

At the same time, there are other aspects of imaging technology that have strong economical implications in other areas besides color communication. The other applications are derived from what is considered the very heart of such a system – the spectral base for color. Contrary to most CAD type systems, the input and output channels are spectral channels – reflectance values either measured or generated which are, for practical purposes, largely device and illuminant independent. This is sometimes referred to as TCITCO (True Color In- True Color Out). The spectral data is by far the most basic characterization of an object’s color. From these spectral values, we derive all the other higher level output forms such as colorimetric values (X,Y,Z, L*,a*,b*,C*H*), output to the monitor in calibrated color (R’,G’,B’), and to the calibrated printer in C’,M’,Y’,K’. By combining the spectral base, colorimetric functions, and an image processor, the color imaging system is a powerful tool for color management.

Color Imaging and Spectral Curves

Spectral input from instruments is not the only method to form this basis of color. The imaging system also relies upon a Curve Generator. This curve generator works in much the same way as color formula prediction in that the unknown color is synthesized from a database of primaries according to the following general formula:

K/S? (predited) = K/S? (primary 1) + K/S? (primary 2) + K/S? (primary 3) . . . + K/S? (substrate)

The initial curve generators in imaging systems were based upon generic primaries. Today the systems have been further developed to allow application-specific primaries to be used for curve generation thereby allowing more realistic and physically achievable color predictions.

New Applications in Imaging

Having briefly described the aspects of imaging that insure TCITCO, the following applications and their benefits to the colorist will be discussed in more detail:

· Integration of spectral-based imaging and color visualization in CAD
· Visualization of Shade Libraries, Production Archiving,
· Calibrated Printing of realistic representations of textiles
· Simulations of standard and user-defined color tolerances
· Visualization of Production Shade-Bands
· Imaging and prediction of colors in multi-step processes

Integration into the CAD World
It has been a dream for many years to provide designers with true-color CAD systems. Such a system would encompass both the graphic and artistic tools of design along with the spectral and colorimetric accuracy of a color management system. As reported in a previous symposium2 designers will soon be able to design products and develop color for these products using spectral-based color management tools and databases.

Visualization of Shade Libraries and Databases
Searching for the right color, or one that matches a new standard, can be a tedious process. Color QC and formulation programs have traditionally provided color search tools, however the results were usually presented to the user in numerical format. While useful to a colorist well trained in colorimetry, those in design and product development need to visualize the result, and preferably in various illuminants, and on the substrate of choice. An imaging system provides such capability by displaying library and archive search results on the substrates required and in order of best to worst. By using the curve generator, a designer can visualize any number of colors in gradients, or variations in lightness, chroma, and hue, between two colors in the library. In this way the designer has available to view and select any of the millions of colors within the gamut of the monitor. Of course by using application-specific primaries for curve generation, the curves are generated using the real-world of substrates and colorants.

Accurate Printing using Spectral-based Print Calibration
In the early days of color imaging, the emphasis was placed upon achieving accurate monitor color. Having achieved optimum performance in this area with current state of monitor technology, the majority of users wanted to print the colors. This was chiefly in the areas of pre-sales, pre-production in development for displaying color trends, a seasonal palette, story boards, and even company shade cards for sales. The process of printing accurate colors was very time consuming requiring mostly trial-and-error using color profiles and shade books. However, spectral-based printer calibrations lead to a more accurate transformation of XYZ to CMYK for that specific printer. In this process, color grids are printed on the substrate of choice and read using a spectrophotometer. A matrix is calculated for the specific printer and substrate and stored as a profile on that imaging system.

Using the 24 standard colors for the Gretag Macbeth Color Checker™, a study3 was conducted using an Epson™ ink-jet printer and several types of paper stock to compare the color quality or accuracy of this method. With this printer on both matte and glossy photo-quality papers, the color differences were calculated between the 24 standards and the printed colors.

For uncalibrated vs. uncalibrated printing, the dE(cielab) values for illuminant D65/10 degree observer are shown in Tables 1-2.

Table 1.    Accuracy of Print Calibration on 24 Color Standards – D65/10 degree

       Table 2.   Accuracy of Print Calibration on 24 Color Standards – D65/10 degree

The results clearly show significant improvement in color quality in reducing the color differences from dE 15.7 to 5.7 average. In a later study4, this method was again evaluated to determine whether the printed colors were considered a visual match by a group of independent subjects. A comparison was also made between the spectral-based print calibration and a well known ICC profiling method available from Kodak™. The results are shown in Figure 1.

Figure 1. Matches considered visually acceptable using two methods of print calibration

While there is still room for further development, many in the industry indicate that the calibrated color is now suitable for some pre-production applications. Substantial savings in time and internal cost can be achieved with this method of calibration and the amount of manual adjustment to a palette of colors is minimized.

Simulations of standard and user-defined color tolerances
Setting realistic expectations on color quality for a particular product can be a daunting experience for the colorist. Those with experience may be able to “feel” their way through and set tolerances usually based upon CMC, or another transformed color space such as CIE 2000. Imaging technology makes it possible for the standard and the tolerance limits in lightness, chroma, and hue to be easily visualized as a grid of seven colors. In this way the colorist use the color generator to visually test possible dE(cmc) tolerances and determine the optimum limit for that particular product.

Visualization of Production Shade-Bands
In the retail and apparel world, the approval of production lots requires a different color model than that of the lab dip process. Specifically, the production lots will have variations in lightness, chroma, and hue according to the variability in dyeing. However to the specifier, these variations must be tightly controlled to prevent color variations from appearing to the consumer at the point of sale. Historically these band definitions were determined by visual inspection and a limit was chosen from among the first dyelots to represent the allowable variation in lightness, chroma, and hue. Subsequent dyelots must fall between these limits to be acceptable, hence the practice of visual shade-banding. Recent work by a larger retailer and a vertical supplier has confirmed that shade bands can be developed visually but stored and implemented digitally by defining in L,C,H the allowable tolerances. This is done by visually approving or rejecting the first series of dyelots, thereby defining numerically an L,C,H band of acceptability. These non-symetrical bands can be applied as tolerances to incoming dyelots, and transmitted via email to the production site for quick approvals of lots in process.
The ability to visualize these bands is a useful tool to confirm that the dyelot in question is within the band, and if not, to what degree it is outside. For marginal dyelots, this allows the colorist to make a decision without delaying production.

Prediction and visualization of color in multi-step processes
Many textiles undergo finishing processes that usually affect the color and appearance. A physical process such as napping will change the substrate structure which affects the appearance and consequently the color. The same is true for yarn dyed knits and wovens. Some finishes are chemical based and cause color change without causing structural changes. These process steps are monitored for color and in most cases it is the experience of the dyer and colorist that must be called upon to make best estimates of these color changes.

By using new imaging tools, a technique is emerging that looks promising in color control in that by using the curve generator it is possible to simulate the finish color based upon historical datasets. In one example, a polyester knit is piece dyed, then processed to produce a very textured napped surface. The manufacturer must match the standard color in finished form, but pass approval on lab dyeings of the unfinished knitted polyester. Additionally, the dyer must be able to produce a formula that will produce the desired finished color by assessing the color of the unfinished fabric. Ideally, the mill would have reliable targets or standards representing each process step. While this is workable on repeat shades, it is guesswork and experience for new colors. Consequently this process could lead to mismatches due to the lack of accurate color matching during these steps.

The prediction process must work in either direction. For a known finished color, the unfinished color may be predicted, and vice versa. A knowledge base is required since this process has been found to be multi-dimensional and non-linear. An analysis of 12 knowns resulted in the following color changes when comparing the unfinished (standard) to the finished color. In this case the changes were physical, not chemically induced.

Table 3. Color Change due to finishing in dE(cmc 2:1) on 12 standard colors (D65/10)

The technique for predicting the color change relies upon two powerful functions within the imaging system. One, as mentioned earlier, is the color generator, or curve generator. Colors may be generated on demand based upon generic or user-defined primaries. Secondly, there must be one or more functions to describe the relationship(s) that exist between the samples representing the various processing steps. The simplest model, and one we are currently pursuing, is to characterize the relationship between the substrates using their respective images as captured by a scanner or camera. This is essentially a “gray-scale” or non-chromatic function. Next, the colorimetric relationship is applied as determined by measurements with a spectrophotometer. By combining the substrate images and their spectral properties it is possible to model the relationship.

At the time of this printing initial results on a small database of samples can be reported and it is expected that additional test samples will be included at the time of the presentation.
It was noted immediately that the color changes were along all 3 color dimensions, and that they were not consistent, or linear. There does however appear to be more consistency among samples of similar chroma and hue. This leads to a prediction model based upon a colorimetric search for the closest match in the database of known, followed by application of the image prediction method as described above.

By using the above 12 pairs, the concept was tested for several colors that were in similar groups such as the Brick Red with Burgundy, Sky Blue with Wisteria, and Deep Forest with Mineral Green. The results of these predictions are given in Table 4 below:

Table 4. Predicted finished color vs. actual measured color

As expected, the results on predictions using Brick Red and Burgundy were better since they are the most similar in L*,C*, and H*. Considering that the finishing changes the color by more than 2 units dE(cmc) the prediction to within 0.7 dE is approaching the acceptable level. It is estimated that on new colors, if the database finds a match to within 5 units dE(cmc) then the prediction of finished (or unfinished) color will be within 0.5 dE(cmc). Additional samples are forthcoming to better determine the accuracy level of this method. Suffice it to say that these results are overall encouraging and with further development, provide another method for dyemills to predict color targets for processing steps.

Conclusions

Color imaging technology has evolved into a very functional toolkit for the colorist working in retail, manufacturing, or dyeing and finishing. The availability of a curve generator and image processing functions provide capabilities that extend beyond the world of color communication and color approvals. With this brief introduction it is clear that there are significant benefits to be realized, and additional application testing is needed to fully achieve results of economic importance.

References

1. Book of papers, AATCC Symposium "Color Innovations 2002, Concepts, Communication, and Control", June 3-4, Raleigh, NC.
2. Ibid
3. Schuman, Les, "A Comparison of Calibrated and Uncalibrated Printing Using the Colorite™ IMProof Print Calibration", October 5, 2000
4. Schuman, Les, "Accuracy in Color Printing using Colorite™, Kodak Color Profiling CME, and PhotoCal RGB print profiling", August 13,2001

The author wishes to acknowledge and thank Malden Mills of Lawrence, MA for color samples for multi-step processes.

Library Index | Home