Continuous Parametric Surfaces for Robust Geometric Modeling

This direction is no longer present in the current build. Its last version stays readable and citable here.

Version 1 of 116 papers21 conceptsfrozen 2026-10-04

In traditional computer graphics and medical imaging, we often represent 3D shapes as discrete collections of parts: voxels in a grid or triangles in a mesh. While intuitive, these representations are inherently "jagged." They lack the smoothness required for high-fidelity physics and create significant hurdles for gradient-based optimization.

A growing body of research proposes a shift from these discrete primitives to continuous parametric surfaces. By representing geometry through mathematical functions—such as B-splines, Non-Uniform Rational B-Splines (NURBS), or Radial Basis Functions (RBFs)—researchers are enabling simulations and reconstructions that are more accurate, differentiable, and computationally efficient.

The Limits of Discretization

When we model a complex biological structure like a platelet or a human face using a piecewise-linear mesh, we approximate a smooth curve with flat segments. This creates two primary problems:

  1. Derivative Noise: Calculating surface normals or curvature requires derivatives. On a discrete mesh, these are often approximated from the angles between flat faces, leading to "noisy" gradients that can destabilize simulations.
  2. Topological Artifacts: During extreme deformations, discrete meshes can fold or fragment. In medical segmentation, voxel-based masks often result in "staircase" artifacts that do not reflect true anatomy.

As Varun Shankar, Grady B. Wright, and colleagues demonstrate in A Study of Different Modeling Choices For Simulating Platelets Within the Immersed Boundary Method, replacing these meshes with continuous parametric models allows for the analytical calculation of surface normals and internal forces. This transition can improve accuracy while being orders of magnitude faster in 3D simulations.

Smoothness as a Prior for Reconstruction

In tasks like 3D face reconstruction, the goal is to infer a surface from 2D images. Traditional methods often produce fragmented or noisy results. By using B-splines, researchers like Zhiyong Feng and Chao Xu in B-spline Shape from Motion & Shading treat the surface as a continuous function with $C^2$ continuity (meaning the surface and its first two derivatives are continuous).

This mathematical smoothness acts as a powerful inductive bias. It ensures that the reconstructed face is naturally organic, eliminating the need for heavy post-process smoothing that might erase fine details. Furthermore, Feng and Xu extended this to Parametric T-Spline Face Morphable Models, which allow for local refinement. Instead of increasing the resolution of the entire mesh, they use quad-tree division to add control points only where high detail is needed, such as around the eyes or mouth.

Differentiability for Optimization and Control

One of the most significant advantages of parametric surfaces is that they are differentiable. In Optimal three dimensional aircraft terrain following and collision avoidance, Tapan Sharma and Paul Williams show that approximating jagged terrain with cubic B-splines provides the smooth gradients necessary for a minimax optimal control solver to function. Without this continuity, the solver would likely fail to converge when encountering sharp edges.

Similarly, in the realm of statistical modeling, Timo Bolkart and Stefanie Wuhrer utilize thin-plate spline mappings in A Groupwise Multilinear Correspondence Optimization for 3D Faces. By optimizing in a continuous 2D parameter space rather than manipulating discrete vertices, they prevent mesh folding and significantly reduce the dimensionality of the optimization problem.

Bridging the Gap to Simulation

A recurring theme in this research is making models "CFD-ready." In engineering and medicine, a 3D model is often just the first step; the next is simulating how blood flows through an artery or how air flows over a wing.

Traditional pipelines require manual, labor-intensive "cleaning" of meshes to make them watertight and smooth enough for Computational Fluid Dynamics (CFD). Work by Jian-Xun Wang and colleagues, such as AortaDiff and HUG-VAS, bypasses this by using NURBS to generate aortic geometries. Because the output is a continuous mathematical surface, it is inherently watertight and smooth, allowing for immediate simulation without manual intervention.

By moving the "source of truth" for geometry from discrete points to continuous functions, these researchers have created a framework where reconstruction, optimization, and physical simulation can finally speak the same mathematical language.

Papers in this version

History & citationVersion 1 · 2026-10-04

No longer present in the current build · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_continuous_parametric_surfaces_for_robus_9fba92ce,
  title = {Continuous Parametric Surfaces for Robust Geometric Modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/continuous-parametric-surfaces-for-robust-geometric-modeling/rdr_9fba92ce572a4f398f6a7e10e79a50b2},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-22},
}

This exact version (version 1, frozen text):

@misc{lacuna_continuous_parametric_surfaces_for_robus_9fba92ce_v1,
  title = {Continuous Parametric Surfaces for Robust Geometric Modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/continuous-parametric-surfaces-for-robust-geometric-modeling/rdr_9fba92ce572a4f398f6a7e10e79a50b2/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-22},
}