Exploring Pipeline of Algorithms for Generating Smooth Meshes From Voxel Data

Abstract This paper focuses on the investigation of a pipeline (choice and sequence) of algorithms for generating smooth triangular meshes from voxel data. Voxel data has become a common place in engineering design and modeling techniques such as topology optimization. A large number of voxels in the design space increases the computational time required for post-processing while inadequate number of voxels leads to improper mesh creation and defects. Thus, it is important not only to choose the optimal voxel sizes and numbers for a given design space but also to smooth appropriately the results to avoid topology and geometry related issues in the downstream processes. There are many different algorithms available for smoothing voxel data and triangular meshes. Each smoothing algorithm has its own advantages and drawbacks. Multiple smoothing algorithms and their application sequence are explored in both voxel space and mesh space. Two classes of metrics are utilized for evaluating the smoothing algorithms - local and global metrics. Local metrics include wavy ridges, bumps, thin holes and thin walls close to one voxel size. The global metric includes the total volume of a mesh. A short review on accurate marching cubes algorithm is also presented as it is a critical step in generating topology preserving mesh. A smoothing pipeline (sequence of selected algorithms) is discussed based on local and global metrics. Results are demonstrated on several examples of voxel data including computational time and volume changes.