Details:
Summary:
While the ultimate test of the quality of a color image reproduction may be the customer's approval, this test does not provide a quantitative basis for the development and improvement of color reproduction systems. Over the past 60 years, there have been numerous attempts to mathematically model 3- and 4-color process printing and proofing systems. Ideally, such a model, when calibrated with measurements of proofed exhibits, should be able to predict the color produced by any given color component (e.g., CMYK) input to the proofing process. Generally, these attempts at developing process color models have not been successful. The shortcomings of these efforts are evidenced by the large numbers of algorithms that are supposed to describe the same or similar reproduction processes. If any of these models gave results that correlated closely with process color reproduction, we could reasonably expect that the others would be only historical curiosities (e.g., the 50+ year old Neugebauer equations are still the basis for the many models of process color). Generally, these attempts at developing process color models have not been successful. The shortcomings of these efforts are evidenced by the large numbers of algorithms that are supposed to describe the same or similar reproduction processes. If any of these models gave results that correlated closely with process color reproduction, we could reasonably expect that the others would be only historical curiosities (e.g., the 50+ year old Neugebauer equations are still the basis for the many models of process color). One reliable, but inelegant, method for determining the relationship between the color component input and the color output (i.e., the transfer function) is the preparation and measurement of proof sheets using a full range of input values component colors. Neugebauer, in a 1942 French patent (#884313), suggested that a table of color values for 1,000 color component inputs would provide an adequate model of a 3-color process. At DuPont, we have used a similar concept to characterize 4-color proofing and printing systems. On well-behaved systems, we measure 7,000 exhibits; on less well-behaved systems (e.g., thermal dye transfer), we use about 15,000 colors. The practicality of this approach is highly dependent on efficient measuring equipment. (Our current equipment takes about 1,000 full-spectrum measurements per hour.) Examination of our data shows that the use of selected input values, instead of an evenly spaced array of input values, allows the reduction of the number of inputs by 50% to 75%. Unfortunately, this selection process depends on the use of a detailed model of each system. Therefore, we must measure 7,000 to 15,000 samples before we can select an optimum number of inputs. When we perform statistical studies where repeated characterizations of the same system are required, this reduced number of measurements can speed the quantitative evaluations of a system.