2019 | vol. 67 | nr. 4 | art. 11

An Image Segmentation Algorithm based on LSM with Stochastic Constraint applied to Computed Tomography Images

Messaouda LARBI, Zoubeida MESSALI, Ahmed HAFAIFA, Abdellah KOUZOU, Tarek Fortaki
Abstract
A new and efficient stochastic level set method based on stochastic formulation, for image segmentation, is presented in this paper. The proposed method can detect objects with weak boundaries using edges and region information as a prior information. The proposed method is characterized by using two different distributions which suit very well to model the background and the object present in the image. Namely, the Gaussian and the Rayleigh distributions are used to model the random variables of the object and the background, respectively. This choice is justified by the behaviour of both the background and the object. Afterwards, an energy function based on a Bayesian rule is minimized to achieve the segmentation. The main goal of our approach is to improve the detection of objects with missing parts. We conduct several experiments using medical Computed Tomography images (CT) to evaluate the performance of the proposed method. The obtained results confirm the superiority of our scheme in terms of precision and robustness in segmenting images with weak borders and lost component, while compared with other state-of-the-art level set approaches.
Keywords: Image Segmentation; Bayesian Estimation; Level Set Method; Active Contours; Otsu's method; CT Images
To cite this article: Messaouda LARBI, Zoubeida MESSALI, Ahmed HAFAIFA, Abdellah KOUZOU, Tarek FORTAKI“An image Segmentation Algorithm based on LSM with Stochastic Constraint applied to Computed Tomography Images”, in Electrotehnica, Electronica, Automatica (EEA), 2019, vol. 67, no. 4, pp. 87-96, ISSN 1582-5175.

 

Download