Publication

Abstract : Hyperspectral image processing is an important area of research nowadays. Since the cameras used for capturing the hyperspectral images is having low spatial resolution, the spectra of observed pixels will be the mixtures of various present in the scene. Thus spectral unmixing aims at estimating the no. of endmembers(reference materials), their spectral signatures and corresponding abundance maps in the captured hyperspectral data. This paper presents performance evaluation and comparative study of statistical and geometrical approaches used for spectral unmixing. The algorithms evaluated are ICA(independent Component Analysis), AVMAX (Alternating volume maximization), SVMAX (Successive volume maximization) and ADVMM (Alternating decoupled volume max-min).The algorithms are implemented and validated on real hyperspectral dataset AVIRIS cuprite data collected over Nevada, U.S in 1997.