Hierarchical Sparse Nonnegative Matrix Factorization for Hyperspectral Unmixing with Spectral Variability

Tatsumi Uezato (1) , Mathieu Fauvel (2) , Nicolas Dobigeon (3, 4, 5)

1 RIKEN AIP - RIKEN Center for Advanced Intelligence Project [Tokyo]
2 CESBIO - Centre d'études spatiales de la biosphère
3 IRIT-SC - Signal et Communications
4 IUF - Institut universitaire de France
5 Toulouse INP - Institut National Polytechnique (Toulouse)

Accounting for endmember variability is a challenging issue when unmixing hyperspectral data. This paper models the variability that is associated with each endmember as a conical hull defined by extremal pixels from the data set. These extremal pixels are considered as so-called prototypal endmember spectra that have meaningful physical interpretation. Capitalizing on this data-driven modeling, the pixels of the hyperspectral image are then described as combinations of these prototypal endmember spectra weighted by bundling coefficients and spatial abundances. The proposed unmixing model not only extracts and clusters the prototypal endmember spectra, but also estimates the abundances of each endmember. The performance of the approach is illustrated thanks to experiments conducted on simulated and real hyperspectral data and it outperforms state-of-the-art methods.

Mots clés: hyperspectral imaging; spectral unmixing; sparse unmixing; endmember variability

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement des images [eess.IV]; Apprentissage [cs.LG]; Analyse de données, Statistiques et Probabilités [physics.data-an]; Vision par ordinateur et reconnaissance de formes [cs.CV]; Intelligence artificielle [cs.AI]

Tatsumi Uezato, Mathieu Fauvel, Nicolas Dobigeon. Hierarchical Sparse Nonnegative Matrix Factorization for Hyperspectral Unmixing with Spectral Variability. Remote Sensing, 2020, 12 (14), pp.2326. ⟨10.3390/rs12142326⟩. ⟨hal-02949164⟩