Sampling and reconstruction with adaptive meshes

Abstract

This paper introduces an approach to visual sampling and reconstruction motivated by concepts from numerical grid generation. We develop adaptive meshes that can nonuni- formly sample and reconstruct intensity and range data. Adaptive meshes are dynamic models which are assembled by interconnecting nodal masses with adjustable springs. Acting as mobile sampling sites, the nodes observe interest- ing properties of the input data, such as intensities, depths, gradients, and curvatures. Based on these nodal observa- tions, the springs automatically adjust their stiffnesses so as to distribute the available degrees of freedom of the re- constructed model in accordance with the local complexity of the input data. The adaptive mesh algorithm runs at in- teractive rates with continuous 3D display on a graphics workstation. We apply it to the adaptive sampling and re- construction of images and surfaces.