Matlab implementation of Nonlinear Extended Blind End-member and Abundance Extraction (NEBEAE) algorithm.
NEBEAE implements a blind hyperspectral unmixing based on the multilinear mixing model (MMM).
In the problem formulation, we include a normalization step in the hyperspectral measurements for the end-members and abundances to improve robustness.
The blind unmixing process can be separated into three estimation subproblems for each component in the model, which are solved by a cyclic coordinate descent algorithm and quadratic constrained optimizations. Each problem is mathematically formulated and derived to construct the overall nonlinear iterative unmixing technique.
We evaluated our proposal with synthetic and experimental datasets from the remote sensing literature (Cuprite and Urban datasets).
The performance is compared with two state-of-the-art unmixing methods based on MMM: (i) Multilinear Mixing Model for Nonlinear Spectral Unmixing (MMMNSU) (end-members initialized by VCA), and (ii) Unsupervised Nonlinear Spectral Unmixing Based on MMM (UNSUBMMM).
The file PlotSyntheticDatabase_Figures2_3.m evaluates the unmixing of the syntetic dataset. As example, the following image shows the estimated abundance maps, and the histogram of the resulting nonlinear interaction level.
The file Table5_Figure5Cuprite.m evaluates the unmixing of the Cuprite dataset. As example, the following image shows the map of resulting the nonlinear interaction level.
The file Table6_Figure6Urban.m evaluates the unmixing of the for Urban Dataset.
The file Table7_Figure7Pavia.m evaluates the unmixing of the for Pavia Dataset.
The file PlotNonlinearVNIR_invivoBrain_Figure8.m unmix the in-vivo dataset (Due to data size restrictions in github, please contact the authors requesting VNIR type images).
| Paper
Nonlinear Extended Blind End-member and Abundance Extraction for Hyperspectral Images
Daniel U. Campos-Delgado1,2,
Inés A. Cruz-Guerrero2
Juan N. Mendoza-Chavarría2,
Aldo R. Mejía-Rodríguez2,
Samuel Ortega3,4,
Himar Fabelo4,
Gustavo M. Callico4
1Optical Communication Research Institute (IICO), Autonomous University of San Luis Potosí, Av. Karakorum 1470, 78210, S.L.P., México
2Faculty of Science, Autonomous University of San Luis Potosí, Av. Parque Chapultepec 1570, 78290, S.L.P., Mexico
3Norwegian Institute of Food Fisheries and Aquaculture Research (NOFIMA), 9019 Tromsø, Norway
4Institute for Applied Microelectronics (IUMA), University of Las Palmas de Gran Canaria, E35017 Las Palmas de Gran Canaria, Spain
Submitted to Signal Processing (Elsevier).
@article{CAMPOSDELGADO2022108718,
title = {Nonlinear extended blind end-member and abundance extraction for hyperspectral images},
journal = {Signal Processing},
volume = {201},
pages = {108718},
year = {2022},
issn = {0165-1684},
doi = {https://doi.org/10.1016/j.sigpro.2022.108718},
url = {https://www.sciencedirect.com/science/article/pii/S0165168422002572},
author = {Daniel U. Campos-Delgado and Inés A. Cruz-Guerrero and Juan N. Mendoza-Chavarría and Aldo R. Mejía-Rodríguez and Samuel Ortega and Himar Fabelo and Gustavo M. Callico},
keywords = {Nonlinear unmixing, Hyperspectral imaging, Multi-linear model}
}