Addressing Spectral Variability in Hyperspectral Unmixing with a Novel Linear Adaptive Additive Mixing Model and an Associated ADMM-Based Approach

Spectral variability often affects remote sensing hyperspectral data, presenting challenges for their unmixing process when considering the traditional linear mixing model, which ignores such an issue. Consequently, more advanced linear mixing models, which address this phenomenon, need to be considered. In this work, an original linear adaptive additive mixing model is proposed to deal with spectral variability. It considers this concern in the spectral part of the manipulated variables. In addition, an associated unmixing algorithm, based on an alternating direction method of multipliers framework, is designed for the proposed model. The usefulness of this algorithm is clearly proven by outcomes of experiments carried out with synthetic data.