Building Color-Management Modules Using Linear Optimization I. Desktop Color System

Abstract

Two kinds of models were derived that predicted spectral reflectance factor of colors formed using an ink-jet printer. One was the spectral Murray–Davies–Yule–Nielsen model in which n-value was assumed to vary as a function of wavelength. The other was based on the Omatsu model in which the path length of light scattering was assumed to vary as a function of wavelength. Model parameters were optimized using a test target of 57 samples consisting of cyan, magenta, yellow, red, green, blue, and black colors varying between white and the maximum ink amount. Average accuracy of an independent data set sampling the printer's color gamut was 4.2 and 3.9 ΔEab*, for the Murray–Davies–Yule–Nielsen and the Omatsu models, respectively. The difference in performance was not significant. The Yule–Nielsen model was selected to build device profiles because of its simplicity in comparison to the Omatsu model. A desktop scanner was colorimetrically characterized using a multiple-linear-regression model to build a concatenated device profile in which digital counts of a scanned photographic reflection print were the input and those of the printer were the output. Because the printer model was analytically noninvertable, the Newton-Raphson and the Simplex iterative methods were evaluated as candidate optimization methods to build 33 × 33 × 33 color look-up tables. These tables were evaluated by comparing a photographic reflection IT8.7/2 target with its printed reproduction. The Simplex method yielded superior results, particularly for colors near the edge or outside of the printer's color gamut. The average ΔEab* error from a profile based on the Simplex method was 5.9 including colors outside of the printer's color gamut.