Skip to Main content Skip to Navigation
New interface
Theses

Segmentation itérative d'images par propagation de connaissances dans le domaine possibiliste : application à la détection de tumeurs en imagerie mammographique

Abstract : The segmentation step constitutes a central point of image interpretation and scene analysis processes. Although the segmentation has already been the subject of numerous studies, it generally remains highly dependent on the nature of images and on their exploitation context. It is worthwhile to notice that the application of different segmentation methods on a given image often produces different results. In order to obtain a reliable and robust segmentation process, we mainly focus on the knowledge contained in the image, and especially on its representation space, aiming at exploiting the maximum of available information, both in terms of its image characteristics and spatial structure. In this context, we propose an approach of image segmentation based on the possibility theory. This approach allows representing image data by taking into account the ambiguity of the information it represents, based on a descriptive knowledge expressed by experts. Furthermore, during the process of pixel-based classification, the proposed approach simulates an important aspect of the human reasoning, the attention focusing, by the progressive integration of information related to the spatial context of pixels. This cognitive process is modeled by the iterative diffusion of the available knowledge, within the neighborhood of image pixels, particularly from possibilistic germs, assumed to represent points of the scene that belongs to a given semantic class with high certainty. Several strategies for both knowledge diffusion and the integration of germs are proposed and evaluated, qualitatively and quantitatively. The overall validation of the proposed approach is performed using a representative set of mammographic images. Obtained results tend to show the superiority of the proposed approach, in terms of accuracy and specificity, by comparison to more classical methods.
Document type :
Theses
Complete list of metadata

https://theses.hal.science/tel-00741996
Contributor : Bibliothèque Télécom Bretagne Connect in order to contact the contributor
Submitted on : Monday, October 15, 2012 - 4:14:29 PM
Last modification on : Saturday, June 25, 2022 - 9:04:19 PM
Long-term archiving on: : Thursday, January 17, 2013 - 11:25:33 AM

File

2012telb0231_Eziddin_Wael.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-00741996, version 1

Citation

Wael Eziddin. Segmentation itérative d'images par propagation de connaissances dans le domaine possibiliste : application à la détection de tumeurs en imagerie mammographique. Traitement des images [eess.IV]. Télécom Bretagne, Université de Bretagne Occidentale, 2012. Français. ⟨NNT : ⟩. ⟨tel-00741996⟩

Share

Metrics

Record views

352

Files downloads

1256