Hardware-Accelerated Reconstruction of Polygonal Isosurface Representations on Unstructured Grids
PG '04 Proceedings of the Computer Graphics and Applications, 12th Pacific Conference
VISSYM'04 Proceedings of the Sixth Joint Eurographics - IEEE TCVG conference on Visualization
Interactive volume isosurface rendering using BT volumes
Proceedings of the 2008 symposium on Interactive 3D graphics and games
Real-Time Isosurface Extraction Using the GPU Programmable Geometry Pipeline
ACM SIGGRAPH 2007 courses
Advanced interactive medical visualization on the GPU
Journal of Parallel and Distributed Computing
Load-balanced isosurfacing on multi-GPU clusters
EG PGV'10 Proceedings of the 10th Eurographics conference on Parallel Graphics and Visualization
From multiple views to textured 3d meshes: a GPU-Powered approach
ECCV'10 Proceedings of the 11th European conference on Trends and Topics in Computer Vision - Volume Part II
Triangulating molecular surfaces on multiple GPUs
Proceedings of the 20th European MPI Users' Group Meeting
Hi-index | 0.00 |
Medical imaging and scientific simulation produce large volumetric datasets, which are often visualized by isosurface extraction and rendering. This extracts individual surfaces from the volume to represent significant boundaries in the volume. The standard isosurface extraction method, Marching Cubes, generates a triangulated approximation of the isosurface one cell at a time.We contribute improvements with respect to algorithmic acceleration and to optimization using programmable hardware on commodity GPUs. We improve algorithmic performance by precomputing topology for each cell and storing this information in the span space structure, avoiding redundant runtime CPU computations in the CPU at runtime.Our principal contribution, however, is to improve GPU-based isosurface acceleration by caching the topology on the GPU in the form of display lists and using vertex programs to perform geometric interpolation. Pascucci[3]and Klein[2] accelerated isosurfaces on tetrahedral meshes by 30%, while Reck[4] achieved 850% using span space techniques. For cubic data, Goetz[1] accelerated Marching Cubes on the GPU but without span space or correct normals.We generate accurate isosurfaces with correct normals for cubic meshes with acceleration at least as good as Pascucci[3], rising to 300% when span space structures are used. Combined with span space acceleration we achieved gains of as much as 1300%, in line with the results of Reck[4] for tetrahedra. Moreover, our method is independent of cell shape and isosurface kernel, and can be applied to higher-order isosurface interpolation. Further acceleration is also expected using improved vertex textures in the next generation of GPUs.