Research
STEM microscopy community uses the notion of spectrum-image since the seminal paper by Jeanguillaume and Colliex in 1989. At a completely different scale, hyperspectral images are widely studied in domains ranging from food monitoring to astronomy. Hyperspectral images represent huge quantities of data and a lot of different techniques have been developped, particularly in the remote sensing domain with the development of satellite imagery.
As theses techniques required advanced mathematics I have a collaboration with the team of Nicolas Dobigeon, professor in the INP-ENSEEIHT, University of Toulouse (http://dobigeon.perso.enseeiht.fr/)
Among the different challenges in data analysis, spectral unmixing aims at providing comprehensive description of the hyperspectral measurements. It consist of extracting the spectral signatures (components) that are characteristics of the main compounds present in the sample and quantifying their respective spatial distribution (abundances)over the image.
We have tested several algorithms : Vertex Component Analysis and Bayesian Linear Unmixing.
https://www.stem.lps.u-psud.fr/spectral-mixture-analysis-eels-spectrum-i...
Full publications list
(Publications listed here present works also not endorsing the STEM group)