Simultaneous model-based segmentation of multiple objects

Abstract

Deformable models are used for the segmentation of objects in 3D images by adapting flexible meshes to image structures. The simultaneous segmentation of multiple objects often causes problems like violation of spatial relationships. Here we present two methods for including prior knowledge into the segmentation process: The first deals with objects sliding with respect to each other, and the second considers a pre-defined minimal distance between neighboring objects. Using this prior knowledge improved segmentation results can be reached.