Hyperspectral Unmixing via Deep Autoencoder Networks for a Generalized Linear-Mixture/Nonlinear-Fluctuation Model

Abstract

Spectral unmixing is an important task in hyperspectral image processing for separating the mixed spectral data pertaining to various materials observed individual pixels. Recently, nonlinear spectral unmixing has received particular attention because a linear mixture is not appropriate under many conditions. However, existing nonlinear unmixing approaches are often based on specific assumptions regarding the inherent nonlinearity, and they can be ineffective when applied to conditions deviating from the original assumptions. Therefore, these approaches are not well suited to scenes with unknown nonlinearity characteristics. This paper presents an unsupervised nonlinear spectral unmixing method based on a deep autoencoder network that applies to a generalized linear-mixture/nonlinear fluctuation model, consisting of a linear mixture component and an additive nonlinear mixture component that depends on both endmembers and abundances. The proposed approach benefits from the universal modeling ability of deep neural networks to learn the inherent nonlinearity of the nonlinear mixture component from the data itself via the autoencoder network, rather than relying on an assumed form. Extensive experiments with numerically synthetic, labeled laboratory-created data and real airborne data, illustrate the generality and effectiveness of this approach compared with state-of-the-art methods.