Free-form deformation of solid geometric models
Thomas W. SederbergScott R. Parry
Introduces free-form deformation, a foundational technique that sculpts arbitrary solid geometric models, parametric surfaces, and polygonal meshes by embedding them in a lattice of trivariate Bernstein polynomial control points with derivative continuity and volume-preservation controls.
A new technique called free-form deformation enables designers to reshape solid geometric models by embedding them in a flexible lattice and then moving control points on that lattice, much as a sculptor works with clay. Solid modeling systems have long relied on simple surfaces such as planes and quadrics, while free-form surface modeling has remained largely separate; the gap has made it difficult to create and edit complex, curved solids inside standard modeling pipelines.
The authors set out to show that a single, simple mapping based on trivariate Bernstein polynomials can deform any solid model—whether defined by constructive solid geometry or boundary representation—while preserving the validity of the underlying representation. They formulated the deformation by first attaching a local coordinate frame to a parallelepiped region, placing a regular grid of control points inside it, and then displacing those points; the new position of any point inside the solid is obtained by evaluating the resulting tensor-product polynomial. The method was demonstrated on polygonal data, implicit surfaces, parametric patches, and complete CSG models, including hierarchical sequences of local and global deformations.
The work shows that free-form deformation applies uniformly to every common surface type and modeling scheme, that derivative continuity across patch boundaries can be maintained by leaving neighboring control points fixed, and that a family of deformations leaves volume exactly unchanged. It further demonstrates that parametric curves and surfaces remain parametric after deformation, that mass-property calculations stay feasible, and that the technique produces usable results for aesthetic shapes, transition surfaces, and some blended features. Professional stylists who tried the method found the sculpting metaphor natural and efficient; one complete telephone handset model was produced in a single design pass after only a few hours of practice.
These capabilities matter because they let engineers and stylists edit free-form solids inside existing solid-modeling systems without switching representations or losing downstream analysis tools. Volume preservation removes a common source of error in packaging and fluid-volume calculations, while the ability to keep surfaces parametric supports continued use of standard intersection and rendering algorithms.
The authors note three practical limits: the method does not yet provide general filleting and blending, local deformations produce planar interfaces with the undeformed region, and operations on the trivariate polynomials are more expensive than the corresponding surface operations. They also observe that the full usefulness of volume-preserving deformations remains to be explored. Further work on display speed, inverse mapping robustness, and integration with commercial fillet routines would strengthen the approach before broad deployment.
No sufficiently relevant recommendations were found.
- Paper: Lucas-Kanade 20 Years On: A Unifying Framework, Simon Baker et al. (2004). This paper builds directly upon foundational continuous spatial transformations and optimization techniques established in early graphics and vision literature to unify subsequent algorithmic variants.
