Vol. 42, Issue 4, pp. 913-924 (2012)

Abstract

A new method for selection of optimal physical primaries is introduced. The reflectance spectra of 1269 matt Munsell color chips are used as a universal physical dataset and the most independent samples are extracted and used as actual primaries. The efficiency of selected primaries is compared with those obtained from principal component analysis (PCA), non-negative matrix factorization (NNMF) and non-linear principal component analysis (NLPCA) techniques. The performances of chosen primaries are evaluated by calculation of root mean square (RMS) errors, the goodness of fit coefficient (GFC) and the color difference (ΔE) values between the original and the reconstructed spectra.