Accueil DE EN ES FR


Advanced Search

Our On-Line PhDs

Submit a Thesis
My Account Register Help

About
Fields
Mathematics and Applications
Information and Communication Sciences and Technologies
Physics, Optics
Materials Science, Mechanics and Mechanical Engineering
Fluid Mechanics and Energy
Chemistry, Physical Chemistry and Chemical Engineering
Life Sciences and Engineering
Earth Sciences and Environmental Engineering
Sciences of Economy, Management and Society
Apprentissage a contrario et architecture efficace pour la détection d'évènements visuels significatifs

Burrus, Nicolas (2008) Apprentissage a contrario et architecture efficace pour la détection d'évènements visuels significatifs. PhD thesis Informatique, Unité Electronique et Informatique ENSTA ParisTech p.164.

Full text available as:

- these-burrus.pdf ( 16133 Kb )
Licence: Copyright

Abstract

To ensure the robustness of a detection algorithm, it is important to get a close control of the false alarms it may produce. Because of the great variability of natural images, this task is very difficult in computer vision, and most methods have to rely on a priori chosen parameters.

This limits the validity and applicability of the resulting algorithms. Recently, by searching for structures for which some properties are very unlikely to be due to chance, the a contrario statistical approach has proved successful to provide parameterless detection algorithms with a bounded expected number of false alarms.

However, existing applications rely on a purely analytical framework that requires a big modeling effort, makes it difficult to use heterogeneous features and limits the use of data-driven search heuristics. In this thesis, we propose to overcome these restrictions by using statistical learning for quantities that cannot be computed analytically. The interest of this approach is demonstrated through three applications: segment detection, segmentation into homogeneous regions, and object matching from a database of pictures. For the two first ones, we show that robust decision thresholds can be learned from white noise images. For the last one, we show that only a few examples of natural images that do not contain the database objects are sufficient to learn accurate decision thresholds.

Finally, we notice that the monotonicity of a contrario reasoning enables an incremental integration of partial data. This property leads us to propose an architecture for object detection which has an “anytime” behavior : it provides results all along its execution, the most salient first, and thus can be constrained to run in limited time.

Item Type:PhD Thesis (PhD)
PhD Supervisor:Bernard, Thierry and Jolion, Jean-Michel
Date:2008
Board of examiners:Bernard, Thierry and Blanc-Talon, Jacques and Collin, Bertrand and Cord, Matthieu and Jolion, Jean-Michel and Merigot, Alain and Moisan, Lionel
Ecole Doctorale:ED 130 INFORMATIQUE, TELECOMMUNICATIONS ET ELECTRONIQUE (EDITE)
Discipline:Informatique
Collection (Fonds):ENSTA ParisTech
Institution:Université Pierre et Marie Curie (UPMC)
Department:Unité Electronique et Informatique ENSTA ParisTech
Subjects:2. Information and Communication Sciences and Technologies
Uncontrolled Keywords:Vision par ordinateur, Méthodes a contrario, Apprentissage statistique, Segmentation, Reconnaissance d’objets, Anytime, Computer vision, A contrario reasoning, Statistical learning, Segmentation, Object detection, Anytime
ID Code:5245
Deposited By:Sophie Chouaf
Deposited On:08 July 2009

Statistiques de consultation

Repository Staff Only: edit this item

© ParisTech 2007 - Réalisé par RILK.com - Graphisme par Winch Communication