Nonlinear hyperspectral unmixing accounting for spatial illumination variability

This paper presents a new supervised algorithm for nonlinear hyperspectral unmixing. Based on the residual component analysis model, the proposed model assumes the linear model to be corrupted by an additive term that accounts for bilinear interactions between the endmembers. The proposed formulation considers also the effect of the spatial illumination variability. The parameters of the proposed model are estimated using a Bayesian strategy. This approach introduces prior distributions on the parameters of interest to take into account their known constraints. The resulting posterior distribution is optimized using a coordinate descent algorithm which allows us to approximate the maximum a posteriori estimator of the unknown model parameters. The proposed model and estimation algorithm are validated on both synthetic and real images showing competitive results regarding the quality of the inferences and the computational complexity when compared to the state-of-the-art algorithms.