The Use of a Selective Database Technique in Order to Recover the Spectra of a Series of Acrylic Paints by the Principle Component Analysis

Abstract

A procedure for an efficient recovering of reflectance spectra of Acrylic paint samples from CIE tristimulus color values is described. By fixing a certain criteria based on color difference value, the proposed technique preliminarily selects a series of suitable samples from a main dataset containing the reflectance values of a series of different Acrylic paint samples, based on the color specifications of given samples. In this way, a series of different databases could be formed around a particular sample. The well-known principal-components linear model was used to recover the spectral data from their corresponding color coordinates by using only 3 basis functions. The surface spectra of a set of 2802 samples are collected for the recovery of the reflectance values of Acrylic paint samples whose tristimulus values were known. The role of the value of color difference for selecting suitable samples is discussed. The recovered spectra achieved by this method show considerable improvements in terms of root mean squarer (RMS) error and goodness- fitting coefficient as well as color difference values under different illuminants as compared to the recovery from the main database.