Computational design of simulated multi-primary display spectral bands for spectral colour reproduction

Abstract

We proposed an optimal combination of primary spectra in six-primary display to reproduce the great majority of spectra from natural objects. We applied normal distribution function to simulate the basic bands of six-primary display, and then we simulated a large data set by least-squares method created from the product of Standard Object Color Spectra (SOCS) database and the irradiance data of Judd et al. [1]. A Good final result was obtained with the averaged root-mean-square error (RMSE) from the simulated sample being 0.0132, demonstrating a significant improvement compared to previous studies. We also compared this method to other multi-spectrum methods (Ajito et al. [2], Long and Fairchild [3], and non-negative matrix factorization (NMF)) for an expanded sample of SOCS data. The results showed our proposed method had an appropriately sized area in the CIE (u′, v′) chromaticity diagram, also with smooth spectral distribution and good reproduction. The spectral radiance distributions can be reproduced well, especially in natural spectral data with the proposed spectral bands.