Vous êtes ici : Accueil » Kiosque » Annonce

Identification

Identifiant: 
Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?

Annonce

6 juillet 2018

Reflectance estimation from multiview RGB images


Catégorie : Post-doctorant


Keywords: Reflectance estimation, Multiview, shading,

Duration: 12 months
 
Start: As soon as possible, applications will be reviewed when they arrive until one candidate is selected
 
Location: Dijon France (Univ. Bourgogne Franche-Comté in the LE2I lab)
 
Subject: Estimating the shape and the reflectance of a scene is a key issue in computer vision and it has many ap-plications such as shape recovery, specularity removal and facilitates classical image processing and com-puter vision tasks such as segmentation or classification. Unfortunately, the estimation of reflectance, lighting and the geometry from RGB images is an underdetermined problem. One single image could have been obtained by a large combination of illumination, shapes and reflectance. Mathematically, there are much more unknowns than equations.
 
To overcome the underdetermination, some recent works have proposed to add some constraints into the optimization process or to increase the amount of data available, for example by combining multiple views of the scene. This post-doc proposal is within this context and is applied to the field of endoscopy. The main objective is to use multiple views of the same stomach tissue to alleviate the underdetermination of the inversion problem and estimate the reflectance of the tissue from RGB images.
 
The work will be carried out within the framework of the French ANR project Emmie which focuses on mul-timodal endoscopy for the detection of inflammatory lesions in the stomach.
 
Recently, we have proposed an optimization method to retrieve the parameters of the dichromatic model from an RGB image using two independent quadratic programming steps. Compared to the state of the art in this domain, our approach has the advantage to transform a complex inverse problem into two paral-lelizable optimization steps that are much easier to solve and do not require a tedious learning step. Anoth-er research team participating in the ANR Emmie project also proposed a method to obtain 2D mosaics by registering the images of a video-sequence acquired during endoscopic examinations. Therefore several views from the same region of the stomach are available.
 
Application:
The candidate is expected to have a PhD degree in a relevant subject area (Image Processing, Applied Math-ematics, Biomedical engineering or Computer Science). Strong analytical/mathematical skills are essential and good programming experience is preferable. It is expected that the candidate has good communication skills, especially in written English.
 
To apply, please email:
•a CV,
•a transcript of your MSc grades/marks (even if incomplete),
•the main paper you wrote during your PhD thesis
•at least one recommendation letter
 
Contact:
•Yannick Benezeth - yannick.benezeth@u-bourgogne.fr
•Franck Marzani- franck.marzani@u-bourgogne.fr
 

Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2018.