A Variational Representation for Efficient Noisy Segmentation

R. Romano and D. Vitulano
Istituto per le Applicazioni del Calcolo "M. Picone"
C. N. R.
Viale del Policlinico 137
00161 Rome, Italy
Tel. +39--6--88470224, Fax +39--6--4404306
E-mail: {romano, vitulano}@iac.rm.cnr.it

Keywords: Image Segmentation, Variational Models, Textures


In this paper we focus on a novel technique VRENS, (Variational Representation for Efficient Noisy Segmentation) to segment images (i.e. containg also textures) under noise. It is based on a hierarchical representation obtained by a combination of the classical weak membrane and a simpler region competition method. VRENS seems to be an interesting algorithm because of its robustness to added gaussian noise along with a low computational cost.