Multiresolution Volume Simplification and Polygonization

Abstract

We propose a multiresolution volume simplification and polygonization algorithm. Traditionally, voxel-based algorithms lack the adaptive resolution support and consequently simplified volumes quickly lose sharp features after several levels of downsampling, while tetrahedral-based simplification algorithms usually generate poorly shaped triangles. In our method, each boundary cell is represented by a carefully selected representative vertex. The quadric error metrics are applied as the geometric error metric. Our approach first builds an error pyramid by bottom-up cell merging. We avoid topology problems in hierarchical cell merging by disabling erroneous cells and penalizing cells containing disconnected surface components with additional costs. Then, a top-down traversal is used to collect cells within a user specified error threshold. The surfacenets algorithm is used to polygonize these cells. We enhance it with online triangle shape optimization and budget control. Finally, we discuss a novel octree implementation which greatly eases the polygonization operations.