Intrinsic Image Decomposition embedded Sparse Spectral Unmixing for Satellite Hyperspectral Images

Abstract

Spectral variability (SV) due to external factors such as atmospheric, illumination and environmental changes is unavoidable in spectral unmixing (SU) of satellite hyperspectral images (HSIs). Considering libraries of a priori acquired spectra is one of the most important approaches to dealing with the problem of SV. SU has recently been applied to hyperspectral imagery, where the goal is to select a limited number of spectral features that can represent each observed pixel well. Intrinsic Image decomposition (IID) can recover the intrinsic reflectance component thus reduce the effect of SV. Based on this, a novel Intrinsic Image Decomposition embedded Sparse Spectral Unmixing (IIDSSU) model is proposed by replacing the original hyperspectral with the intrinsic reflectance component, which is independent of changes in external imaging conditions. Experimental validation is performed using satellite HSI from the Yellow River Delta region. The results show that the robustness and superiority of the unmixing results can be efficiently enhanced by the proposed IIDSSU.