Analysis of Hyperspectral Images Based on PCA

Abstract

Recent computational models of color vision demonstrate that it is possible to achieve exact color constancy over a limited range of lights and surfaces described by a low-dimensional linear model. For smooth reflectance spectra the different spectral bands have a significant degree of correlation. Any spectral reflectance distribution can be approximated to a specified degree of accuracy as a weighted sum of basis functions. 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 data sets such as rocks, leaves, Munsell chips, etc. The characteristics of the spectra will be important to understand the spectral properties of the object reflectance and the representation of spectral images by linear models. In this study sets of hyperspectral images have been analyzed by principal-component analysis (PCA) method for spectral encoding.