Building a metric of color reproduction difference by combining multiple observers in a modular online experiment

Abstract

A web-hosted online experiment was previously developed to find the visual difference between four reproduction gamuts using direct magnitude estimation (Proc. IS&T 29th Color and Imaging Conf, 2021:317–322). In order to increase the size of the data set, but without overburdening observers, a modular approach was adopted. The original methodology was therefore extended across 10 linked sub-experiments to make comparisons between some 36 gamuts, which were designed to exhibit a variety of different gamut shapes, contrast ratios, and substrate colors within the constraints of a desktop display. In addition to each set of test images, a common normalization set was included in all sub-experiments in order to adjust each observer's choice of modulus to a global average observer, and thus combine the results into a larger data set. Finally, an interval scale was inferred from the normalized magnitude data using a categorical judgment approach to calculate scale values. The fitted data revealed a power function close to a square-root between the interval and magnitude scales.