Shape Analysis Lecture 6: Surface Curvature Fundamentals

Added:

Discrete Differential Geometry and Discrete Curvature

While classical differential geometry relies on the assumption of smooth, infinitely differentiable surfaces, modern shape analysis in computer science primarily deals with discrete representations like triangle meshes and point clouds. Applying classical smooth curvature formulas (such as the second fundamental form) directly to these piecewise-linear surfaces is mathematically problematic because derivatives do not exist at edges and vertices. Discrete Differential Geometry (DDG) serves as a critical alternative paradigm. Instead of merely approximating smooth curvatures through discretization, DDG reconstructs geometric theories from the ground up for discrete structures. It defines discrete analogues of curvature—such as the angle defect for Gaussian curvature and cotangent weights for mean curvature—specifically designed to preserve global topological and geometric invariants (like the Gauss-Bonnet theorem) exactly on the mesh. This discrete viewpoint is often more computationally robust, mathematically rigorous, and structurally consistent for digital shape processing than attempting to apply classical continuous differential geometry to discrete data.

Multivariable Calculus: Proficiency with surface parametrization, tangent planes, and computing surface normal vectors.
Linear Algebra: A strong understanding of eigenvalues, eigenvectors, and symmetric bilinear forms, which are crucial for analyzing the shape operator.
First Fundamental Form: Familiarity with the metric tensor, arc length, and how distances and angles are measured on a parameterized surface.
Differential Geometry of Curves: An understanding of curvature, torsion, and the Frenet-Serret frame for one-dimensional space curves.
Theorema Egregium: Exploring Gauss's landmark theorem which proves that Gaussian curvature is an intrinsic property of a surface.
The Gauss-Bonnet Theorem: Understanding the profound connection between a surface's local geometry (curvature) and its global topology (Euler characteristic).
Minimal Surfaces: Studying surfaces with zero mean curvature, including physical applications like soap films and mathematical modeling.
Discrete Differential Geometry and Computer Graphics: Applying these curvature concepts to discrete triangle meshes for 3D shape segmentation, processing, and rendering.
8.6K views189likes1:11:15@justinmsolomonOriginal Release: 2021-04-15

The second fundamental form is a symmetric bilinear form on the tangent plane of a surface, defined as II(v, w) = -v · dn(w), where dn is the differential of the Gauss map (shape operator). This form encodes how the surface bends in space and has eigenvalues called principal curvatures (κ₁, κ₂) whose product gives Gaussian curvature (K = κ₁κ₂) and average gives mean curvature (H = (κ₁+κ₂)/2). Gaussian curvature distinguishes surface types: positive K indicates elliptic (bowl-shaped) surfaces, negative K indicates hyperbolic (saddle-shaped) surfaces, and zero K indicates parabolic (cylindrical) surfaces. Mean curvature measures how the surface interacts with its normal vector.