Computational protocol: Quality Partitioned Meshing of Multi-Material Objects

Similar protocols

Protocol publication

[…] The input 3D multi-material classification map is often obtained by multi-domain segmentation or classifiers using normalized graph cut operating on a voxellized 3D reconstructed image of the MMO [,]. Softwares such as AsymSeg and SymSeg [,] implemented and publicly available from VolRover [] and Segger [] can be used to generate such material classification maps. Generally, if a face is shared by two voxels with the same material ID i, then the face will be removed and the two voxels are clustered into the same i material region. Otherwise, the face will be a part of the partition mesh Pt and will be partitioned into two different material regions. Unfortunately, the partition meshes {Pt}t=1N generated by this simple way usually suffer from the following problems: (a) non-manifold problem, including non-manifold vertices and edges; (b) tiny component problem, like isolated and extremely tiny independent material regions. Our proposed algorithm is: Fix the tiny component problem.Fix the non-manifold problem.Partition the voxel classification map and generate the multi-material mesh.Optimize the generated partition mesh.Fix the tiny component problem.Fix the non-manifold problem.Partition the voxel classification map and generate the multi-material mesh.Optimize the generated partition mesh.More details of each step are provided in sub-sections 4.1, 4.2, 4.3 and 4.4, respectively. When we process our input solid domain, we perform steps 1 and 2 repeatedly and in that order, i.e., we first fix the tiny component problem, and then fix the non-manifold problem. If the non-manifold fix step creates tiny components, we repeat the tiny component fix steps as necessary. […]

Pipeline specifications

Software tools VolumeRover, Segger
Application cryo-EM