Reproduce “A Region-based Randers Geodesic Approach for Image Segmentation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Da Chen, Jean-Marie Mirebeau, Huazhong Shu, Laurent D. Cohen
The paper "A Region-based Randers Geodesic Approach for Image Segmentation" by Chen et al. introduces a variational image segmentation model that reformulates region-based active contour energy into a Randers geodesic problem. This approach integrates regional homogeneity and edge features directly into the path cost, enabling robust segmentation in images with weak edges or noise and demonstrating improved accuracy compared to existing methods.
Key insights
- Region-based active contour energy can be mathematically transformed into a minimal path problem within a Randers geometric framework, integrating regional homogeneity directly into the geodesic metric definition.
- The use of an asymmetric Randers metric, where the vector field is derived from the region-based energy gradient, provides a robust mechanism to guide geodesic paths even in the absence of strong or reliable edge information.
- Confining contour evolution to a dynamically adjusted tubular domain, combined with the efficient FM-ASR solver, enables a computationally feasible and theoretically convergent shape evolution scheme for segmentation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.