Fast processing spectral discrimination for hyperspectral imagers based on interferometry

Abstract

Hyperspectral imagers based on interferometry associate with each pixel of the image the spectrum calculated with the Fourier transform of the measured interferogram. This class of hyperspectral imagers is intrinsically faster than imagers based on optical filters and on dispersive means when the noise is dominated by detector noise. This speed advantage could be hampered by the large computing power necessary to extract the spectral content of the image from the large number of data acquired by the camera. We have realized a real-time algorithm to discriminate different spectra of the pixels of a scene directly from the acquired interferograms. The technique showed very good discrimination of pixels illuminated by narrow band radiation. This algorithm is based on principal component analysis and could be implemented directly in the camera processor to discriminate spectra in the image in real time using factorization with singular value decomposition.