Reach For the Spheres: Tangency-aware surface reconstruction of SDFs
DescriptionSigned distance fields (SDFs) are a widely utilized implicit surface representation that has applications in various fields such as computer graphics, computer vision, and applied mathematics. Despite their frequent use, traditional methods such as Marching Cubes and its variants often overlook fundamental properties of SDFs, resulting in mesh reconstructions that exhibit oversmoothing and feature loss. To address this, we propose a novel method based on a key insight: each SDF sample corresponds to a spherical region, and the surface represented by the discrete SDF has critical properties with respect to each sphere. Leveraging this understanding, our method formulates an energy that gauges the degree of violation of tangency constraints by a proposed surface. We then employ a gradient flow that minimizes our energy, starting from an initial triangle mesh that encapsulates the surface. The algorithm yields superior reconstructions, surpassing the quality of previous methods, even with sparsely sampled SDFs. Our approach provides a more nuanced understanding of SDFs and offers significant improvements in surface reconstruction.
Event Type
Technical Papers
TimeTuesday, 12 December 20239:30am - 12:45pm
LocationDarling Harbour Theatre, Level 2 (Convention Centre)
