A new angle on hue scaling

Abstract

Hue scaling is a common measure of color appearance in which a hue is decomposed into the perceived proportions of the opponent primaries red vs. green and blue vs. yellow. The method has been important for deriving the putative opponent response functions of human color vision, but can be non-intuitive and poses well-known problems for estimating and analyzing proportions. We explored an alternative scaling procedure in which text labels for the four opponent primaries were placed along the cardinal compass points of a circular display and observers then set the needle to rate the hue of the test stimuli. This allows the scaling to be based on a more explicit color-similarity rating with the underlying proportions given by these ratings. We tested a sample of observers (UNR students) with both tasks, estimating the hue of the same stimuli by varying the compass angle or rating the primary percentages. The stimuli were 30 chromatic angles at 12-deg intervals and constant contrast in a scaled MacLeod-Boynton space. Observers repeated each task twice in counterbalanced order. We compare the two tasks in terms of within-observer variability to assess the reliability of the judgments, and also between observer differences to explore how the estimated opponent response functions are affected by the task. The task-dependent differences are important because they place limits on inferences that can be drawn about color appearance from hue scaling. We also report preliminary data using the compass task to assess hue scaling in two groups field-tested in India.