"Scale Space Meshing of Raw Data Point Sets."

Julie Digne, Jean-Michel Morel, Charyar-Mehdi Souzani and Claire Lartigue

Abstract :

This paper develops a scale space strategy for the meshing and segmentation of complete raw data points sets. The scale space is based on the intrinsic heat equation (mean curvature motion, MCM). A simple iterative scheme implementing MCM directly on the raw points is described, and a mathematical proof of its consistency with MCM given. Points evolved by this MCM implementation can be trivially backtracked to their initial raw position. A consequence is that the reversible MCM scheme permits to first orient, and then mesh reliably a raw textured surface. The accuracy gain is demonstrated on archaeological objects by comparisons with other meshing methods. The obtained discrete 3D scale space also complies with its traditional role: It permits to segment the original raw surface into ridges and valleys computed at coarse scales, and to draw the meaningful inflexion lines on the raw meshed surface.