General Deformable Model Approach for Model-Based Reconstruction
Résumé
In this paper we propose a general approach for performing both model-based reconstruction and segmentation. We extract geometric models from 3D datasets by deforming a reference model. Classical approaches either perform registration of a template by evaluating global transformations or use the deformable models framework to apply free-form deformations. Some authors have proposed to combine both approaches. In our approach we introduce an a priori knowledge about the geometry and the topology of the objects to recover. Furthermore we show in this paper the link between computing a global transformation on a model and applying an external force field to a deformable model. We thus deduce a computer-efficient hybrid deformation scheme allowing to control the number of degrees of freedom of the deformable models through a single parameter. We then present a general framework integrating registration, hybrid and free-form deformations that performs robust reconstruction and segmentation. We insist on the geometric quality of the retrieved models.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...