The color of pigment mixtures is essential for the appearance of an artwork. In this work, we propose a physically-based pipeline to model such mixtures according to the optical properties of the pigments using the Kubelka–Munk turbid media theory. More precisely, we estimate the scattering and absorption coefficients of pigments from color charts applied on a four-level contrast paper, measured with a hyperspectral imaging device. Based on this multi-tonal background, we aim to improve the traditional black-white form of the Kubelka–Munk optical model. We compare this method with two other solutions of the Kubelka–Munk model, namely single-constant simplification and non-linear optimization of the general formulation. Given the estimated optical properties, we digitally simulate the reflectance for mixtures of up to 5 base pigments. Afterwards, we evaluate the spectral and colorimetric accuracy of these digital simulations against a reference target that was physically created by mixing the base pigments at controlled concentrations. Furthermore, we test the performance of our approach in reconstructing the color of a more spatially complex stimulus—a painting. Results show that the four-level background brings an improvement to the recovery of pigment optical properties with respect to the two-level black-white method.
