Hardware-Accelerated Interval Tree Index for Marching Cubes

This paper proposes a novel hardware-accelerated interval tree index for marching cubes,which exploits the massively parallel architecture of modern GPU for accelerating isosurface extraction.In the pre-computation stage,our method partitions volume data into sub-volumes and then indexes voxels based on its range interval in each sub-volume totally on the GPU.In the runtime stage,this index is used to parallelly locate and process active voxels which are intersected by the given isosurface value in marching cubes.We illustrate our method by using various volume data,in which it outperforms all other known GPU-based marching cubes.The computation speed of our method increases up to 4~10 times in comparison with existing GPU benchmark algorithms.