Skip to Main content Skip to Navigation

Détection de petits objets dans une image en utilisant les techniques de super-résolution

Abstract : This thesis deals with the detection of small objects in an image using super-resolution (SR) techniques. Image reconstruction using a SR method consists in producing a high-resolution (HR) image, from several low-resolution (LR) images obtained either from several cameras, or from a video sequence shot with one single video camera. Obtaining a HR image requires two steps: registration of the LR images on a common grid, and the estimation of the HR image using a data fusion approach.

Therefore, this manuscript presents two parts. The first main section deals with detection and registration methods, and the second one with multi-frame SR restoration techniques. About the first part, several methods have been evaluated: a frequency-domain registration method using phase correlation principle on the one hand, and a speck detection method based on a MAP estimator in a Bayesian framework on the other hand. In the second part, a new SR method using a hierarchical Markov model for the HR image in the Bayesian framework is proposed. This new approach, which is based on the idea that the HR image is made of homogeneous regions, provides not only an image of good quality, but also a segmentation result of the HR scene.
Document type :
Complete list of metadatas
Contributor : Fabrice Humblot <>
Submitted on : Saturday, January 7, 2006 - 9:54:42 PM
Last modification on : Wednesday, October 14, 2020 - 3:56:52 AM
Long-term archiving on: : Saturday, April 3, 2010 - 8:58:23 PM


  • HAL Id : tel-00011319, version 1



Fabrice Humblot. Détection de petits objets dans une image en utilisant les techniques de super-résolution. Interface homme-machine [cs.HC]. Université Paris Sud - Paris XI, 2005. Français. ⟨tel-00011319⟩



Record views


Files downloads