ICC.2:2017 is a revision to the next-generation colour management specification iccMAX that introduces new support for colour appearance processing. iccMAX includes a built-in colour appearance model IccCAM, together with a rich programming environment, and support for spectral data, material channel connections, BRDF and processing elements that make it possible to functionally encode any appearance model. ICC.2:2017 introduces many new capabilities, including the ability to provide environment variables which allow parameters such as image statistics or viewing conditions to be passed to the transform at run-time. ICC.2:2017 supports a wide range of colour appearance computations within the colour management workflow.
The performance of Color Prediction Methods CAT02, KSM2, Waypoint, Best Linear, MMV center, and relit color signal are compared in terms of how well they explain Logvinenko & Tokunaga [1] asymmetric color matching results. In their experiment, given a Munsell paper under a test illuminant, 4 observers were asked to determine (3 repeats) which of 22 other Munsell papers made the least-dissimilar match under a match illuminant. Given this data, we address the following four questions. Question 1: Are observers choosing the original Munsell paper under the match illuminant? If they are, then the average (12 matches) color signal (cone response triple or XYZ) made under a given illuminant condition should correspond to that of the Munsell paper's color signal under the match illuminant. Computation shows that in 274 of the 400 cases, the relit color signal is close to the mean color signal of the matches. Question 2: How do algorithm predictions compare to the average observer prediction of the actual color signal of the relit paper? The Wilcoxon signed-rank test shows that KSM2, Waypoint, and Best Linear perform equally, and that both slightly outperform the observer average, which, in turn, significantly outperforms CAT02, and MMV (metamer mismatch volume) center. Question 3: Which method most closely predicts the observer average? We found that the color signal of the relit reflectance is a better predictor of the average observer than Best Linear, which in turn is marginally better than Wpt and KSM2, both of which outperform CATO2 and MMV center. Question 4: Do the observers agree with one another? Using a leave-one-observer-out comparison shows that individual observers predict the average matches of the remaining observers somewhat better than the relit color signal, which in turn slightly outperforms Best Linear, Wpt and KSM2, which then all significantly outperform CATO2 and MMV center.