A fast voxelization algorithm for trilinearly interpolated isosurfaces

Abstract

In this work, we propose a new method for a fast incremental voxelization of isosurfaces obtained by the trilinear interpolation of 3D data. Our objective consists in the fast generation of subvoxelized isosurfaces extracted by a point-based technique similar to the Dividing Cubes algorithm. Our technique involves neither an exhaustive scan search process nor a graph-based search approach when generating isosurface points. Instead an optimized incremental approach is adopted here for a rapid isosurface extraction. With a sufficient sampling subdivision criteria around critical points, the extracted isosurface is both correct and topologically consistent with respect to the piecewise trilinear interpolant. Furthermore, the discretization scheme used in our method ensures obtaining thin - one voxel width - isosurfaces as compared to the one given by the Dividing Cubes algorithm. The resultant subvoxelized isosurfaces are efficiently tested against all possible configurations of the trilinear interpolant and real-world datasets.