Research on Reconstruction of Spectral Reflectance Based on Principal Component Analysis

Abstract

Traditional color reproduction technology based on the Metamerism principle, the disadvantage is that different observer condition leads to different color appearance.To fulfill the color consistency, the spectrum reflectance of the object color sample need to be reconstructed. The principal component analysis makes use of the linear combination of a few principal components to reconstruct the spectral reflectance of sample. This paper analyzes the 31*31 matrix of Munsell spectral data by the principle component analyze method and achieves the principal component for spectrum reflectance. The numbers of principal components are identified as six by discussing the variance contribution rate. Spectral reconstruction of four Munsell testing samples makes use of first six principal components, which has met the accuracy requirements. Research shows that the reconstruction of spectral accuracy decreased when training samples and testing samples belong to the different database.