Level-of-Detail Independent Voxel-Based Surface Approximations

A research project by Robbin Marcus
Conducted at Geometric Computing Group at the Master Game and Media Technology program
Under supervision of Amir Vaxman
Mar 2017
Thesis

We present a voxel-based surface representation that contains level of detail (LOD) independent surfaces. We approximate the surface in every voxel by storing scalar data in every corner. The representation is close to the original surface in terms of contour information. In order to preserve the contour as accurately as possible, we introduce a new error metric that operates on the voxels. By simplifying voxels that contain the least amount of contour information, we achieve a voxel structure with low memory requirements. Our error metric correlates the level of detail with the amount of error of the surface approximation. We can create a model to a specific level of detail by constraining either maximum memory requirements, or the maximum allowed error of the approximation, with respect to the original model.

Computer graphics, geometry processing, voxel rendering, mesh simplification.