We solve two drawbacks of the classical minimal path models. Firstly, the curvature term was removed in the model in order to transform the problem into finding distance map associated to a source point. We reintroduce the curvature term to the Eikonal PDE framework to find curvature regularised minimal paths. Moreover, in the context of segmentation, existing metrics depend only on object boundary information. We propose a new Finsler metric involving the region-based similarity/dissimilarity measure.

This presentation is part of Minisymposium “MS43 - Variational Image Segmentation: Methods and Applications”
organized by: Jack Spencer (University of Liverpool) .

Authors:

Da Chen (University Paris Dauphine, PSL Research University)

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

Keywords:

active contour, geodesic, image segmentation, partial differential equation models