Images et Dynamique

Abstract : The core of this research work concerns the estimation of the dynamics displayed on an image sequence, in particular on satellite acquisitions of the atmosphere and the ocean. The chosen approach is based on the optimal coupling of a numerical model, describing the observed system, and of the images. The mathematical methods are based on data assimilation technics. The model is named ``Image Model'', as it describes the temporal processes creating the evolution of the images. Its design is based either on heuristics or on simplified physical laws, as for instance the shallow water equations that are used for describing the satellite acquisitions of the ocean surface. The document concerns two main types of data assimilation methods: the variational technics, and in particular 4D-Var with or without model error, and the filtering technics, and in particular the ensemble Kalman filter. In each context, the mathematical components of the assimilation system are described and their impact is analyzed. A focus is given on the definition of observations from image data, in order to take into account the displayed structures, and on the definition of the observation operators.
Document type :
Habilitation à diriger des recherches
Complete list of metadatas

Cited literature [97 references]  Display  Hide  Download

https://hal.inria.fr/tel-01245112
Contributor : Nathalie Gaudechoux <>
Submitted on : Wednesday, December 16, 2015 - 5:01:57 PM
Last modification on : Friday, May 25, 2018 - 12:02:03 PM
Long-term archiving on : Saturday, April 29, 2017 - 4:58:15 PM

Identifiers

  • HAL Id : tel-01245112, version 1

Collections

Citation

Isabelle Herlin. Images et Dynamique. Modélisation et simulation. Université Paris 6, 2015. ⟨tel-01245112⟩

Share

Metrics

Record views

383

Files downloads

294