Skip to Main content Skip to Navigation

Conception de métaheuristiques d'optimisation pour la segmentation d'images. Application aux images biomédicales

Abstract : Image segmentation generally is the important step in any automated image analysis system, such as autonomous vehicle navigation, object recognition, etc. All these subsequent tasks, such as feature extraction, object detection, and object recognition, depend on the quality of segmentation. One of the fundamental weaknesses of current image segmentation algorithms is their inability to adapt the segmentation process to different kinds of images.
The metaheuristics appeared in the eighties. These global optimization algorithms are stochastic and can be applied to any problem, at the condition it is formulated as a mono-objective or multiobjective optimization problem. They are inspired from an analogy with physics (simulated annealing, microcanonical annealing), with biology (evolutionary algorithms) or with ethology (ant colonies, particle swarms). They also can be extended, particularly to multiobjective optimization.
In order to design a segmentation system that allows to have good segmentation results on different kinds of images, we formulate the segmentation as: a mono-objective then a multiobjective optimization problem.
In the mono-objective approach, we adapt several metaheuristics to the segmentation problem. Then, an application to the brain magnetic resonance images (MRI) is performed. This adaptation allows to compare the different metaheuristics in terms of complexity, convergence speed, adaptability and solution reproducibility.
Afterwards, we propose a multiobjective optimization approach to solve the image segmentation problem. In this context, we develop three adaptive methods: the first is based on aggregation of criteria, the second is based on a non-Pareto approach and the third is based on a Pareto approach. Finally, we consider the case of brain ventricle space segmentation and we apply the different approaches to sane and pathologic MRI.
Document type :
Complete list of metadatas

Cited literature [121 references]  Display  Hide  Download
Contributor : Amir Nakib <>
Submitted on : Friday, August 1, 2008 - 4:36:16 PM
Last modification on : Wednesday, September 4, 2019 - 1:52:09 PM
Long-term archiving on: : Thursday, June 3, 2010 - 5:36:05 PM


  • HAL Id : tel-00308789, version 1



Amir Nakib. Conception de métaheuristiques d'optimisation pour la segmentation d'images. Application aux images biomédicales. Informatique [cs]. Université Paris XII Val de Marne, 2007. Français. ⟨tel-00308789⟩



Record views


Files downloads