Accelerating surface remeshing through GPU-based computation of the restricted tangent face

Abstract

High-quality mesh surfaces are crucial for geometric processing in a variety of applications. To generate these meshes, polyhedral remeshing techniques truncate Voronoi cells of the original surface and yield precise intersections, but the calculation is complicated. Some methods apply auxiliary points to construct Voronoi diagrams to simplify these techniques, thereby extracting co-planar facets to approximate the original surface. However, extracting these approximate facets from the constructed Voronoi diagram makes it inefficient and non-parallelizable. To this end, we propose an efficient GPU method for manifold surface remeshing, where the restricted tangent face (RTF) is utilized to approximate the original surface. By intersecting the pre-clipped Voronoi cell with the tangent plane, this method directly calculates the RTF of each point without any auxiliary points or traversing Voronoi cells. Moreover, to restrict the movement of points, we introduce a projection method based on the KNN strategy, where each point is projected onto the triangular facet in the original surface. Owing to the independence and non-interference of the RTF computation and projection of each point, our method is implemented in parallel on the GPU. Experimental results on various mesh surfaces demonstrate the superior performance of our method in the viability, effectiveness, and efficiency.