Detection and Correction of Errors in Psychophysical Color Difference Munsell Re-renotation Dataset

Abstract

The Munsell dataset holds a prominent position in the field of color science. This dataset describes large color differences covering a wide color gamut, making it highly valuable for the development of color models. Currently, the widely used version is the Munsell Renotation, which is the second version of the dataset. In this paper, we analyze the third version, known as the Munsell Re-renotation, identify significant errors within it, and provide corrections for obvious typos. We propose a novel method for detecting nonuniformities, utilizing the L1-STRESS measure and the proLab uniform color space (UCS). Our findings demonstrate that the revised version of the Munsell Re-renotation dataset achieves significantly better consistency with established UCSs compared to the original Munsell Re-renotation data. Additionally, we discuss modifications of the STRESS measure for data with unknown scales. Unlike previous modifications, the proposed measure, STRESSgroup, is identical to the classic STRESS measure when the scales are the same.