Skip to Main content Skip to Navigation
Theses

Approximation de l'Information Mutuelle basée sur le développement d'Edgeworth : application au recalage d'images médicales.

Abstract : Mutual Information (MI) is considered as the most common similarity measure in the context of intensity-based image registration. This measure is well-known for its ability to perform tri-dimensional multimodal medical image registration. However, MI's estimators suffer from variance, bias and lead to high computational complexity. During this PhD thesis, we dealt with some statistical tools called cumulants in order to build novel approximations of MI based on Edgeworth expansion. This expansion allows one to approximate a probability density in terms of cumulants. The estimate of these approximations as similarity measure was analyzed in terms of performance on both synthetic and real data, on rigid and non-rigid medical images registration tasks. A comparison with classical estimators of MI was also performed.
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00632128
Contributor : Mathieu Rubeaux <>
Submitted on : Thursday, October 13, 2011 - 3:19:48 PM
Last modification on : Thursday, January 14, 2021 - 11:17:03 AM
Long-term archiving on: : Saturday, January 14, 2012 - 2:27:57 AM

Identifiers

  • HAL Id : tel-00632128, version 1

Collections

Citation

Mathieu Rubeaux. Approximation de l'Information Mutuelle basée sur le développement d'Edgeworth : application au recalage d'images médicales.. Traitement du signal et de l'image [eess.SP]. Université Rennes 1, 2011. Français. ⟨tel-00632128⟩

Share

Metrics

Record views

687

Files downloads

4732