Spectral Reconstruction on the Basis of Several Samples and Principal Components Analysis

Abstract

One important problem of colorimetry is that how to determine the spectral characteristics of colour samples with the help of some known parameters. If we take the spectral reflection features of a big amount of sample sets and use a linear algebraic-statistical method, the principal component analysis, we can determine that handful of parameters whose linear combination helps to reconstruct the spectral distribution of the elements of the original sample sets. The average vector given by this method and the first couple of eigenvectors can reproduce the elements of the samples with a good approximation, therefore, in case of a large set of colour samples, it is enough to weigh the first few base vectors to reconstruct the samples.