Direct Computing of Surface Curvatures for Point-Set Surfaces

Accurate computing of the curvatures of a surface from its discrete form is of fundamental importance for many graphics and engineering applications. The moving least-squares (MLS) surface from Levin [Lev2003] and its variants have been successfully used to define point-set surfaces in a variety of point cloud data based modeling and rendering applications. This paper presents a set of analytical equations for direct computing of surface curvatures from pointset surfaces based on the explicit definition from [AK04a, AK04b]. Besides the Gaussian parameter involved in the MLS definition, these analytical equations allow us to conduct direct and exact differential geometric analysis on the point-set surfaces without specifying any subjective parameters. Our experimental validation on both synthetic and real point cloud data demonstrates that such direct computing from analytical equations provides a viable approach for surface curvature evaluation for unorganized point cloud data.