RESEAUX DE NEURONES EN TRAITEMENT D'IMAGES
- Des Modèles théoriques aux Applications Industrielles -

Gilles Burel 1, 2, 3
2 Lab-STICC_UBO_CACS_COM
Lab-STICC - Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance, UBO - Université de Brest
Abstract : The interest of neural networks in the image and signal processing field is
addressed. This work has been performed inside a research laboratory of an
industrial group. So, we focus on realistic problems, but without neglecting
theoretical research.

In a first part, we prove the
interest of neural networks as a source of inspiration for conception of new
algorithms. We propose, for instance, an original structure for prediction,
and new algorithms for Vector Quantization. The properties of existing
algorithms are also clarified from a theoretical point of view, and methods
for automatic tuning of their parameters are
proposed.

Then, we show that neural networks can solve a
lot of applications of industrial interest. For various image and signal
processing problems (segmentation, pattern recognition, data compression,
blind separation of sources, etc), we show that a neural solution can be
developped at low cost.
Complete list of metadatas

Cited literature [34 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00101699
Contributor : Gilles Burel <>
Submitted on : Wednesday, September 27, 2006 - 7:20:06 PM
Last modification on : Thursday, December 19, 2019 - 1:22:56 AM
Long-term archiving on: Tuesday, April 6, 2010 - 1:15:49 AM

Identifiers

  • HAL Id : tel-00101699, version 1

Citation

Gilles Burel. RESEAUX DE NEURONES EN TRAITEMENT D'IMAGES
- Des Modèles théoriques aux Applications Industrielles -. Traitement du signal et de l'image [eess.SP]. Université de Bretagne occidentale - Brest, 1991. Français. ⟨tel-00101699⟩

Share

Metrics

Record views

1731

Files downloads

28666