Joint Hyperspectral Video Demosaicing and Demixing

When:
30/07/2020 all-day
2020-07-30T02:00:00+02:00
2020-07-30T02:00:00+02:00

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : LISIC, EA 4491
Durée : 3 years
Contact : matthieu.puigt@univ-littoral.fr
Date limite de publication : 2020-07-30

Contexte :
During the last decades, infrared/visible imaging and then multispectral (MS) imaging allowed great breakthoughts, e.g., in industrial or environmental engineering.

The more recent development of hyperspectral (HS) cameras–observing the same image at several hundreds or even thousands wavelengths–makes possible to imagine new observation systems for which novel data processing techniques–at the frontier between image processing and machine learning–must be proposed.

In the context of this Ph.D. thesis, we are particularly interested in HS videos. They provide time sequences of HS data cubes (big data). However, for the sake of miniaturization and of maintaining hardware costs, these cameras do not necessarily acquire all the information they are supposed to sense. Post-processing called “demosaicing” is then necessary to reconstruct the data cube observed at each time instant. Moreover, in each pixel of each image of the HS video, the observed spectrum can be considered as a mixture of spectra of materials present in the pixel.

Sujet :
Within the framework of this Ph.D. thesis, we wish to estimate the whole spectra of all the materials, from partially observed video sequences, in order to perform HS video demosaicing. Several issues such as the mass of data or the spectral variability, will be investigated, ,

From an application point of view, we are interested in monitoring natural, human, or industrial activities. In particular, we will use such HS cameras to monitor coastal or marine fauna.

Profil du candidat :
Prospective applicants should hold a Master degree in Signal/Image Processing, in Machine Learning, in Applied Mathematics, or in any related discipline. Applications from candidates with a good background in (non-negative) matrix/tensor factorization, deep learning, optimization, with excellent programming skills (e.g., in Matlab, Python, C and/or C++) are particularly encouraged.

Applicants are expected to show good communications skills, both written and oral. In particular, speaking fluently in French or English is required. Writing in English is mandatory.

Candidates are requested to send a resume, transcripts from their last year of Bachelor to their last year of Master (if available), as well as two reference letters (or contact details of two referees).

Formation et compétences requises :
Prospective applicants should hold a Master degree in Signal/Image Processing, in Machine Learning, in Applied Mathematics, or in any related discipline.

Adresse d’emploi :
The recruited Ph.D. student will be working in the new antenna of the LISIC laboratory, located in Saint-Omer, Northern France. This antenna is dedicated to MS and HS imagery, with already 3 Ph.D. students and 1 post-doc researcher working in this field.

Document attaché : 202003272316_Joint_demosaicing_demixing_PhD_thesis_2020.pdf