Appropriate Number of Principal Components for Spectral Reflectance Representation of Natural Dyes

Spectral data which give valuable information، are usually represented with a high dimension according to the accuracy of instrument. During the last decades، applying mathematical methods such as Principal Component Analysis (PCA) has shown that spectral data has limited number of principal vectors. In the present study، the appropriate numbers of PC vectors for representing the spectral reflectance of natural dyes were investigated applying a set of 629 wool samples dyed with natural dyes. GFC، RMSE and color differences under two illuminants were considered to evaluate the magnitude of error. Experimental results showed that it is possible to represent the spectral data of natural dyes with almost 5 principal components with precious accuracy (GFC=0. 999). In addition، the best results were obtained when the principal vectors are extracted from the same data set (natural dyes) in comparison to other sets such as dyed polyesters or 1269 Munsell set.