Optimal spatial scale for local region-based active contours

Djamal Boukerroui (1)

1 Heudiasyc - Heuristique et Diagnostic des Systèmes Complexes [Compiègne]

Recently, there has been a reinvestigation of the use of local statistics by the image segmentation community in a variational framework. Locality is generally defined by means of an isotropic spatial kernel. These recent studies show a better behavior of local models on images with strong intensity inhomogeneities. Thus local segmentation models have become very popular. Very little work, however, has been carried out to address the problem of the choice of the local kernel's scale (or bandwidth). The present contribution falls within this context. Specifically, we investigate the problem of the estimation of a single scale (i.e. pixel independent) for local parametric region segmentation models. The approach is generic and straightforward. The method is demonstrated on synthetic and realistic simulations of ultrasound images with intensity inhomogeneities.

Mots clés: Minimization; Kernel; Level set; image segmentation community; intensity inhomogeneities; isotropic spatial kernel; local parametric region segmentation models; local region-based active contours; local statistics; optimal spatial scale; realistic simulations; synthetic simulations; ultrasound images; variational framework; Active contours; Estimation; Image segmentation; variational techniques; Nonhomogeneous media; Bandwidth selection; local active contours

Domaines: Traitement des images [eess.IV]

Djamal Boukerroui. Optimal spatial scale for local region-based active contours. IEEE International Conference on Image Processing, Oct 2014, Paris, France. pp.4393-4397, ⟨10.1109/ICIP.2014.7025891⟩. ⟨hal-01170961⟩