Speaker: | John E. Lavery, Mathematics Division, Army Research Office |
Title: | L1 Splines |
Time: | 3:00-4:00 pm |
Place: | Blocker 628 |
A new class of piecewise polynomial splines, called L1 splines, that are free of the extraneous oscillation seen in conventional splines and that preserve the shape of irregular data well has arisen. The minimization principle for L1 splines is based on the L1 norm rather than the conventional L2 norm. We describe the advantages of L1 splines for univariate and bivariate geometric modeling in Cartesian, polar and spherical coordinate and outline analytical results and numerical issues. Algorithms for L1 splines are adaptions of linear and nonlinear programming algorithms. Non-iterative domain decomposition can be used with these algorithms. Connections with solutions of partial differential equations in L1-based function spaces, with haptics and with pattern recognition will be outlined.
Last revised: 10/16/06 By: christov@math