We propose a dynamic scheme for minimal path computation and tubular centerline extraction. The main contribution lies at the construction of a coherence-penalized Riemannian metric during the geodesic distance computation, dependently of local tubular geometry and an external scalar-valued reference feature map. The geodesics associated to the proposed metric favour to pass through a set of vessel segments with low variations of the feature map, thus can avoid the short branches combination problem.

This is poster number 41 in Poster Session

Authors:

Da Chen (University Paris Dauphine, PSL Research University)

Laurent D. Cohen (University Paris Dauphine, PSL Research University)

Keywords:

geodesic, image segmentation, nonlinear optimization, partial differential equation models, retinal vessel segmentation