Fair Six-Point Subdivision on Euclidean and Riemannian Surfaces

Technical Summary: Fair Six-Point Subdivision on Euclidean and Riemannian Surfaces

Problem Statement
Interpolatory subdivision schemes generate smooth curves from control polygons by iteratively inserting new vertices while retaining existing ones. While the classical four-point Deslauriers–Dubuc (DD) scheme (specifically the Dyn–Gregory–Levin or DGL scheme at tension $\omega=1/16$) is widely used, it produces limit curves with only $C^1$ regularity. This low regularity often results in highly oscillatory curvature profiles, creating tangible artifacts in downstream processes like surface-normal computation and offset-curve generation. Although fairness (smoothness of curvature) is often imposed via post-processing, this paper investigates whether fairness is inherently built into specific classical subdivision rules. Furthermore, the paper addresses the extension of these fairness properties to Riemannian manifolds of constant curvature (the sphere $S^2$ and hyperbolic plane $H^2$), where standard linear subdivision rules do not apply directly.

Methodology
The paper employs a combination of variational calculus, symbolic computation, and numerical analysis to analyze the six-point Deslauriers–Dubuc scheme and its Riemannian analogue.

1. Local Variational Identity: The author establishes a local identity on a single six-point configuration. They fix the four neighboring half-integer insertions via degree-5 Lagrange interpolation from the six reference points. They then define a discrete energy functional penalizing the variation of second differences (a linearization of curvature variation). They prove that the unique minimizer of this energy is exactly the six-point DD midpoint.
2. Manifold Extension: The paper utilizes the "log–exponential" lift framework (Wallner and Dyn; Rahman et al.). The scheme maps six neighboring points to the tangent plane at the geodesic midpoint of the edge, applies the linear six-point mask, and maps the result back to the manifold via the exponential map.
3. Curvature-Variation Modeling: The author derives the full nonlinear Euler–Lagrange equation for the curvature-variation energy $\int (\kappa'_g)^2 ds$ on space forms of constant curvature $K$. They linearize this equation to obtain a two-point model for geodesic curvature profiles, yielding trigonometric solutions on the sphere and hyperbolic-trigonometric solutions on the hyperbolic plane.
4. Numerical Diagnostics: The paper implements the schemes on Euclidean, spherical, and hyperbolic domains. It uses difference-ratio diagnostics to estimate critical Hölder exponents and computes a consistent discrete curvature-variation energy to compare the "fairness" of the four-point, six-point, and eight-point schemes.

Key Contributions
The paper makes five specific contributions, distinguishing between new results and consolidated accounts of prior art:

1. Local Variational Identity (Theorem 1): The author proves that on a six-point configuration with Lagrange-pinned neighbors, the six-point DD midpoint is the unique minimizer of a discrete second-difference variation energy. This provides a structural "fairness reading" of the DD weights, linking them to the minimization of curvature variation in the small-deflection regime.
2. Regularity Analysis: The paper provides a consolidated account of the regularity for the four-, six-, and eight-point DD schemes, confirming critical Hölder exponents of $\approx 2$ ($C^1$), $\approx 2.830$ ($C^2$), and $\approx 3.55$ ($C^3$) respectively. It validates these via numerical difference-ratio diagnostics.
3. Euler–Lagrange Equation on Space Forms: The paper derives the full nonlinear Euler–Lagrange equation for curvature-variation energy on surfaces of constant curvature $K$, retaining all variation terms. It presents the linearized model and its closed-form solutions, identifying the geometric resonance at the antipodal distance on the sphere.
4. Manifold Scheme Implementation: The paper explicitly identifies the geodesic-midpoint log–exponential lift as the established Riemannian Deslauriers–Dubuc analogue (citing Wallner [43]). It provides self-contained implementations for $S^2$ and the Poincaré disk ($H^2$), an elementary proximity bound on the sphere, and a precise explanation of why the ambient-norm argument fails on the hyperboloid.
5. Benchmarking and Diagnostics: The paper presents a benchmark suite comparing the four-, six-, and eight-point schemes. It validates the manifold implementations against known convergence and regularity theory and demonstrates that the six-point scheme balances fairness, locality, and robustness better than its four-point and eight-point counterparts.

Results

• Fairness: Experimental results show that the six-point scheme produces significantly fairer curvature profiles than the four-point DGL scheme. On closed polygons, the discrete curvature-variation energy of the six-point scheme is 5 to nearly two orders of magnitude lower than that of the DGL scheme.
• Regularity: The six-point scheme achieves $C^2$ regularity (critical Hölder exponent $\approx 2.830$), one class higher than the $C^1$ DGL scheme, without the wider support and potential ringing artifacts of the eight-point scheme.
• Manifold Behavior: The manifold scheme preserves the full Euclidean regularity ($C^2$) on both the sphere and hyperbolic plane. Proximity analysis confirms the scheme satisfies the second-order proximity condition required for convergence.
• Null Result on Turning Angles: The paper documents a "cautionary null result": measuring half-edge turning angles on sampled circles fails to detect the curvature-dependent term predicted by the linearized model at order $O(h^3)$. This is because the $K$-independent chord–arc defect enters at the same order, masking the signal. The curvature effects are instead visible through proximity deviations.

Significance and Claims
The paper is modest in its claims regarding novelty, explicitly attributing the construction of the manifold scheme, its convergence, and its regularity to prior work (Wallner, Dyn, Rahman, et al.). The significance of this work lies in:

• Providing an exact local variational interpretation for the six-point DD scheme, characterizing it as a minimizer of discrete curvature variation.
• Offering explicit, self-contained implementations and a proximity bound for the scheme on constant-curvature surfaces.
• Demonstrating through empirical benchmarks that the six-point scheme offers an optimal trade-off between fairness (curvature smoothness), locality (support size), and robustness (handling non-uniform data) compared to the four-point and eight-point alternatives.
• Clarifying the geometric consistency of curvature-variation models on space forms, including the placement of resonances and the limitations of certain diagnostic measurements.

The paper concludes that while the six-point scheme is not a new construction, its identification as a fairness-minimizing rule and its validated performance on Riemannian manifolds provide a robust foundation for high-quality geometric design in both Euclidean and curved spaces.