Kubelka-Munk model of full-gamut oil colour mixing

Abstract

Kubelka-Munk theory (K-M) describes the spectral interaction of light with thin permeable media such as paint [1]. It has found widespread use as an accurate model of colour mixing. Usage often involves characterising a set of base paints1 by measuring the concentrations and reflectance spectra of standardised mix samples, deriving absorption and scattering spectra for each base paint, and using these spectra to predict the colour of new mixes. This paper demonstrates an efficient, accurate application of K-M to a limited palette of five Winsor & Newton oil paints applied in an opaque or alla prima2 method, achieving an average error of 1.49 ΔE00 with 33 mixes. A mix sampling technique is proposed such that the mixed colours comprising the K-M dataset are minimal in quantity, and well-distributed throughout the colour gamut accessible with the base paints, or “full-gamut”. K-M concentration is specified by proxy of base paint mass; accounting for the different oil absorptions and mass densities of the base paints was not found to give any improvement. Model generality was confirmed by holdout cross-validation.