Influence of the noise model on level set active contour segmentation

Abstract

We analyze level set implementation of region snakes based on the maximum likelihood method for different noise models that belong to the exponential family. We show that this approach can improve segmentation results in noisy images and we demonstrate that the regularization term can be efficiently determined using an information theory-based approach, i.e., the minimum description length principle. The criterion to be optimized has no free parameter to be tuned by the user and the obtained segmentation technique is adapted to nonsimply connected objects.

Coupling Image Restoration and Segmentation: A Generalized Linear Model/Bregman Perspective

“…We extend the EF noise model introduced to image segmentation by Chesnaud et al (1999); Martin et al (2004), and Lecellier et al (2010) by integrating image-restoration tasks, such as TV-inpainting and deconvolution. We first present the GLM formulation and show its flexibility in coupling image segmentation and restoration.…”

“…For an introduction to the EF in image segmentation, we refer to the works of Goudail et al (2003), Martin et al (2004), and Lecellier et al (2010) Lecellier et al (2010), in the sense that it is a one-parameter canonical EF with the identity function as its sufficient statistic (cf. Lecellier et al (2010) for the definition of the sufficient statistics vector).…”

A Level Set Method Based on Bayesian Risk for Textured Image Segmentation

“…(2(i, j)), (7)  y (i, j) = sin ( 2(i, j)), (8) International Journal of Machine Learning and Computing, Vol. 7, No.…”

“…The energy functional is designed to obtain a stationary global minimum; thus the energy functional has a unique convergence state, the evolution algorithm is invariant to the initialization, and level set function can set an appropriate termination criterion. Martin et al [8] proposed a level-set active segmentation based on the maximum likelihood estimation to improve the segmented results for several different noise models and showed that the regularity term could be efficiently determined by using the minimum description length (MDL) principle. They assume that noise can be described by members of the exponential family, such as Gaussian, Gamma, Poisson, or Bernoulli distribution.…”

Mammography segmentation with maximum likelihood active contours

“…There is a strong dependence between the segmentation accuracy and the PDF of the noise present in the image (Martin et al, 2004). In this work 1 , we improve the mammography segmentation accuracy by adapting the Gamma distribution, offering two tunable parameters.…”