Real time reconstruction of volumes from very large datasets using CUDA
Lucian Petrescu, Anca Morar, Florica Moldoveanu, Victor Asavei · International Conference on System Theory, Control and Computing · 2011
This article presents a memory efficient implementation of the Marching Cubes algorithm using NVIDIA's CUDA technology. The algorithm can handle datasets that are normally too large for current hardware by splitting the initial volume into several smaller subvolumes while minimizing extra computations caused by subvolume overlapping. Moreover, our approach is scalable, making it easy to benefit from additional computational resources.