Radially-symmetric Compactly-supported Image Modeling

Adrian Hauswirth

Semester Project

Section of Electrical and Electronical Engineering, EPFL

July 2012

Abstract

The goal of this project was to implement an image interpolation algorithm that uses cardinal basis functions constructed from Wendland functions. Using this approach, we experimentally confirmed theoretical results about approximation orders and approximation behavior of these special radial basis functions. Furthermore, we examined the effect of different parameters on the quality of the approximation and showed that interpolation is possible with parameters that imply only a modest computational cost. Finally, we compared the radial basis method with the separable polynomial B-spline model and the experiments indicate that RBF-interpolation provides more favorable results in some cases.