Combination of spectral unmixing algorithms for the fusion of the hyperspectral and multispectral data with unknown spectral response function

Abstract

In this paper Least Square-Non Negative Matrix Factorization Spectral Unmixing Combination (LS-NNMF-SUC) is presented for the fusion of hyperspectral (HS) and multispectral (MS) data. The observed HS and MS images are respectively the spatial and the spectral degradations according to the sensor characteristics of the High spatial-resolution HS (HHS) image, which is reconstructed based on the high spectral information of Low spatial resolution HS (LHS) image represented by endmembers and high spatial information of High spatial-resolution MS (HMS) image represented by abundances. In this work, the proposed algorithm deals with practical remote sensing situation, where the spectral relationship between the observed HMS image and the estimated HHS image is unknown. As a result, a Spectral Unmixing Combination (SUC) diagram based on Least square (LS) and Non-negative Matrix Factorization (NNMF) Spectral Unmixing is developed, in which the one loop NNMF and LS Spectral Unmixing is performed on the MS and HS images sequentially. The spatial spread transform matrix of the sensor observation model is used to produce the matched abundances of the LHS image, in order to unmix the later. Simulation results performed on HYDICE and AVIRIS data demonstrate the efficiency of the proposed fusion algorithm.