Factors Impacting Consistent Color Appearance Prediction
Elena Fedorovskaya and Robert Chung, Rochester Institute of Technology;
David Hunter and Pierre Urbain, Pilot Marketing Group; Don Hutcheson, HutchColor
Achieving consistent color appearance (CCA) despite varying printing conditions is an important goal in print color reproduction. The question becomes: “Is there a set of rules that influence consistent color appearance?”, or, in other words, “Can we model consistent color appearance for image sets?”
In the previous study (TAGA, 2018), 37 color characterization datasets were created whereby 7 of them varied in gamut sizes; the rest of them (30), having the same gamut volume, varied in in-gamut colors. In addition, all of them differed by 3 ∆E00 at the 95th percentile between adjacent datasets . Images, each consisting of a three-scene hardcopy cluster, were printed in accordance to the designed datasets, and psychometric tests were conducted to evaluate CCA between the image sets using rating and rank order procedure. We found a significant correlation between relative CCA interval scores and the measured ∆E00 at the 95th percentile under varying printing conditions. Specifically, by controlling hues of primary and secondary colors, gray balance, and tonality, the printing conditions enable consistent color appearance despite of changes in gamut volume.
In the present study, we plan to use different test images representing larger range of scenes and a softcopy psychometric testing. The three-scene hardcopy cluster is limited in size, scene variety, and takes a considerable time to conduct the psychometric test per observer. In the present study, the images recommended by the CIE TC 8-16 will be used as a single-scene and a montage, and evaluations will be performed by soft proofing using an ISO 12646 and ISO 14861 compliant color monitor with uniformity correction and suitable color management software.
To model consistency of color appearance, we will consider two approaches for quantifying color differences: (a) device-based differences in color distributions between characterization datasets, and (b) image-based differences in color distributions between sampled image pixels. The first method assumes that all colors (1,617 colors in the color characterization target) are equally important. The second method assumes that only the colors in the images matter. We are interested to find out if the image-based color differences can be a better predictor for the CCA than the device-based color differences.
By preparing test images and establishing psychometric testing procedures in a soft proofing environment, we plan to test the following hypotheses:
- There is a significant difference in how color differences are sampled and represented and their correlation with CCA interval scores.
- Psychometric relative CCA interval scores correlate better with the combined device-based color differences and image-based color differences than the device-based color differences along.
We plan to explore several sampling techniques to represent image-based color distributions, including Bicubic interpolation and clustering, as examples.
 Fedorovskaya, Elena, Chung, Robert, Hunter, David, Urbain, Pierre, and Hutcheson, Don, “Defining Consistent Color Appearance for Print Images,” TAGA Proceedings, 2018