A mixed model of active geodesic contours with gradient vector flows for X-ray microtomography segmentation

Laurence Guillot (1) , Emmanuel Le Trong (2) , Olivier Rozenbaum (2) , Maïtine Bergounioux (1) , Jean-Louis Rouet (2)

1 MAPMO - Mathématiques - Analyse, Probabilités, Modélisation - Orléans
2 ISTO - Institut des Sciences de la Terre d'Orléans

The structural characterization of weathered building stones of historical monuments can be achieved with a powerfull imagering technique, X-ray microtomography. It requires however a carefull extraction of each phase constituting the sample from the raw gray-level images obtained (the segmentation process). This contribution presents an original method of segmentation of such images that combines active contour models driven by gradient vector flows with a morphological preprocessing, alternate sequential filters. Preliminary results on high resolution, structuraly complex images are presented and compared to more classical approaches.

Mots clés: Segmentation; Alternate Sequential Filtering (ASF); microtomography; active contours

Domaines: Traitement des images [eess.IV]

Laurence Guillot, Emmanuel Le Trong, Olivier Rozenbaum, Maïtine Bergounioux, Jean-Louis Rouet. A mixed model of active geodesic contours with gradient vector flows for X-ray microtomography segmentation. 2009. ⟨hal-00267007⟩