Chess Download Center

Toolbox ELMM

This code allows to perform the unmixing of hyperspectral images, accounting for endmember variability. We use an Extended Linear Mixing Model (ELMM). Given a hyperspectral data cube, a set of reference endmembers (extracted by any algorithm, or available a priori), and a few parameters, the algorithms returns the abundance coefficients, the scaling factors of the model, and pixel-dependent endmember matrices. Several spatial regularizations can be enforced on the abundances and/or scaling factors.

We use the Extended Linear Mixing Model (ELMM) presented in [1] to unmixing. This model assumes the mixing process is linear, but considers that endmembers are no longer reduced to a single spectral signature, but that they can vary in each pixel of the image. The variations are allowed in the form of scaled versions of reference endmembers, which approximately models spectral variability induced by changing illumination conditions in every pixel. Intrinsic variability of the material can also be captured. In addition, spatial regularizations on the abundances and the scaling factors of the model can be enforced. A detailed presentation and validation of the model and algorithm can be found in:

L. Drumetz, M. A. Veganzones, S. Henrot, R. Phlypo, J. Chanussot and C. Jutten, "Blind Hyperspectral Unmixing Using an Extended Linear Mixing Model to Address Spectral Variability," in IEEE Transactions on Image Processing, vol. 25, no. 8, pp. 3890-3905, Aug. 2016.

The paper and the supplementary materials can be found here :

Please fill the form before downloading the data. This will help us to communicate new experiments and keep you informed about our work. After filling the form, a link will be sent to the e-mail address entered.

Name: Mail: Affiliation: Country: