Abstract
A spectroscopic image analysis based on a subspace method requires orthogonal basis functions which can represent the objective spectrum by the linear combinations. But, It is difficult to realize filter functions corresponding to the spectral sensitivity of orthogonal basis functions. Then we propose a new algorithm to synthesize orthogonal basis functions from broad-band filters using a generalized inverse matrix. Our system consists of a few image sensors and different broad-band filters and does not use scanning process. The resultant spectroscopic images are compared, with the spectroscopic image obtained by conventional method and some components in an image are classfied.
