|M.Sc Student||Kaplansky Lotan|
|Subject||A Level-Set Approach for Curve Propagation on Surface Meshes|
|Department||Department of Electrical Engineering||Supervisor||Professor Ayellet Tal|
|Full Thesis text|
The use of active contour models for a variety of image-based applications has been a lively research topic for the last 20 years. Among the most prominent advances in this field was the introduction of the higher-order level-set framework, which provides a robust, implicit method for curve propagation on 2D images (”snakes”), or surface propagation on 3D images (”balloons”), while automatically handling issues like curve parameterization and topology changes as an inherent part of the flow process. This work presents an extension of the level-set framework from planar, 2D image-based curve flows, to the more general setting of curve evolution on non-planar, freeform surface meshes. Unlike previous approaches for surface curve evolution, our method works by extending the differential operators used in most level-set flows to work directly on the mesh triangulation, while properly adhering to the surface geometry. This results in a parameterization-free, coordinate-free framework, which retains all the benefits of the level-set formulation while remaining general enough to accommodate almost any curve propagation rule. By utilizing curve flows which relate to either the geometrical or external properties of the mesh surface, the algorithm’s versatility is demonstrated in a variety of shape-analysis applications: mesh segmentation improvement, local constriction detection, and texture-based mesh segmentation.