University of Georgia

Mathematics Department Colloquium

John Lavery

Mathematical and Computer Sciences Division

Army Research Office

Research Triangle Park, NC

April 27 1999

Shape-preserving, Multiresolution, Piecewise-polynomial Geometric Modeling without Human Interaction


Abstract: The Applied Analysis Program at the Army Research Office supports research in mathematical analysis for advanced solid materials for structures, armor and sensors, soil and granular materials, fluid flow for chem/bio defense and diesel combustion, photonic bandgaps, nonlinear dynamics for chaotic communication devices, inverse scattering for landmine detection, data fusion and modeling of irregular surfaces.

User-input-free, shape-preserving interpolation and approximation are long-standing goals in geometric modeling. We discuss here a new class of polynomial interpolating and approximating splines that preserve the shape both of smooth data and of data with abrupt changes in magnitude or spacing. The theoretical framework for these splines is straightforward and complete. They have no ad hoc components: One convex nonlinear minimization principle works for all cases (with only a "regularization" parameter needing to be chosen). These splines do not require constraints, penalties, a posteriori filtering or interaction with the user. Univariate and multivariate cases and cubic and higher-degree splines are treated in a unified framework. Finally, these splines can be implemented using efficient interior-point methods for convex nonlinear optimization.