The study of linear model for spectral images

Reflectance spectra of hyperspectral images of the natural scenes are supposed to represent the real world better than any certain classes of natural and man-made spectral reflectance. The question is how the low-dimensional linear model which can be used to approximate Munsell reflectance samples will perform on spectral images and how many parameters are required? To answer these questions, in this paper a statistical analysis of the spectral data sets of spectral images has been applied based on low-dimensional linear modelling. Principal Component Analysis (PCA) technique has been used and sets of reflectance have been reproduced using different numbers of basis functions. The reconstructed spectra have been evaluated and compared with the original spectra. The results show linear models are dependent upon the data sets and small number of basis functions can be used to represent spectral images.