Compactly-Supported Smooth Interpolators for Shape Modeling with Varying\n Resolution

Daniel Schmitter, Julien Fageot, Anaïs Badoual, Pablo García-Amorena, Michaël Unser

Abstract

In applications that involve interactive curve and surface modeling, the intuitive manipulation of shapes is crucial. For instance, user interaction is facilitated if a geometrical object can be manipulated through control points that interpolate the shape itself. Additionally, models for shape representation often need to provide local shape control and they need to be able to reproduce common shape primitives such as ellipsoids, spheres, cylinders, or tori. We present a general framework to construct families of compactly-supported interpolators that are piecewise-exponential polynomial. They can be designed to satisfy regularity constraints of any order and they enable one to build parametric deformable shape models by suitable linear combinations of interpolators. They allow to change the resolution of shapes based on the refinability of B-splines. We illustrate their use on examples to construct shape models that involve curves and surfaces with applications to interactive modeling and character design.

Introduction

Computer graphics and medical imaging often struggle with a fundamental trade-off in shape representation. Discrete models, such as meshes, are easy to refine and offer interpolatory vertices (where points lie directly on the shape), but lack inherent smoothness. Conversely, continuous models like NURBS provide mathematical smoothness and local control but are typically approximating. In approximating models, control points "hover" outside the object, making precise interactive manipulation non-intuitive for designers.

Furthermore, representing common geometric primitives like spheres or tori often requires complex rational functions. This paper introduces a "best of both worlds" framework: a smooth, continuous representation where control points sit exactly on the boundary, and resolution can be increased locally as needed.

Methodology

The proposed framework is built upon Exponential B-splines. Unlike standard polynomial B-splines, these can perfectly reproduce trigonometric and hyperbolic functions—the essential building blocks for circles, ellipses, and complex surfaces.

Half-Integer Shifts

A key technical hurdle is that smooth, compactly-supported interpolators cannot be created using only integer shifts of B-splines. The authors overcome this by employing half-integer shifts (placing basis functions at $n/2$ intervals). This provides the necessary degrees of freedom to satisfy the interpolation condition ($\phi(k) = \delta[k]$) while maintaining smoothness and local support.

A pinched torus modeled with interpolatory control points lying directly on the surface.

Pre-Refinement and Refinability

While standard B-splines are refinable, interpolators typically lose their interpolatory property upon refinement. The authors introduce a pre-refinement step: a pre-filter converts interpolatory control points into a B-spline basis before applying standard refinement iterations. This allows the model to "borrow" the powerful refinement properties of B-splines while maintaining an intuitive interpolatory interface for the user.

Results and Applications

The framework was evaluated through the reproduction of various geometric primitives and complex mathematical surfaces.

Geometric Primitives

The system successfully reproduced shapes such as the torus, helicoid, and pinched torus. By using specific root vectors (e.g., $\alpha = (0, \frac{4i\pi}{M}, -\frac{4i\pi}{M})$), the model can exactly represent complex forms like the Roman surface.

Character Design Case Study

To demonstrate practical utility, the authors applied the framework to a T-Rex character model. A designer can start with a low-resolution "rough" shape and iteratively add detail to specific areas, such as the bones of the foot, without affecting the rest of the model.

Iterative refinement of a T-Rex foot model, demonstrating the transition from initial design to detailed geometry.

Discussion and Limitations

The primary advantage of this approach is direct manipulation: because the basis functions are interpolatory, moving a control point moves the surface exactly at that point. The minimal support of the interpolators ensures local control, preventing unintended deformations in distant parts of the model.

However, these interpolators do not satisfy the convex hull property because the basis functions are not entirely positive. While the shape is not guaranteed to stay within the bounds of its control points, this is often an acceptable trade-off for the benefits of interpolation in interactive design. Additionally, while experimental results strongly support the invertibility of the system's matrix ($\mathbf{A}_\alpha$), a formal mathematical proof remains an open challenge.

Conclusion

This work provides a rigorous mathematical bridge between interpolatory modeling and the refinement capabilities of B-splines. By enabling smooth, local, and intuitive shape control, the framework offers a powerful tool for interactive modeling and character design. Future work may extend these principles to real-time rendering and high-quality image resampling.

Summary updated

Abstract

In applications that involve interactive curve and surface modeling, the\nintuitive manipulation of shapes is crucial. For instance, user interaction is\nfacilitated if a geometrical object can be manipulated through control points\nthat interpolate the shape itself. Additionally, models for shape\nrepresentation often need to provide local shape control and they need to be\nable to reproduce common shape primitives such as ellipsoids, spheres,\ncylinders, or tori. We present a general framework to construct families of\ncompactly-supported interpolators that are piecewise-exponential polynomial.\nThey can be designed to satisfy regularity constraints of any order and they\nenable one to build parametric deformable shape models by suitable linear\ncombinations of interpolators. They allow to change the resolution of shapes\nbased on the refinability of B-splines. We illustrate their use on examples to\nconstruct shape models that involve curves and surfaces with applications to\ninteractive modeling and character design.\n
BibTeX
@article{schmitter2017compactly,
  author = {Daniel Schmitter and Julien Fageot and Anaïs Badoual and Pablo García-Amorena and Michaël Unser},
  title = {{Compactly-Supported Smooth Interpolators for Shape Modeling with Varying\textbackslash\{\}n Resolution}},
  journal = {Graph. Model.},
  year = {2017},
  doi = {10.1016/j.gmod.2017.11.001},
}
Figure