2 of 8

Decimation quality: simple sphere example

Decimate sphere from 20K triangles in 10 times till 2K triangles:

VTK

MeshLib

Ideal mesh

Many near degenerate triangles

3 of 8

Decimation quality: complex example

Original mesh �with 1.1M triangles:

4 of 8

Decimation quality: complex example

Decimate in 100 times till 11K triangles:

VTK

MeshLib

Big holes

Uniform mesh

5 of 8

Mesh subdivision

VTK implementation is not reliable: it can work with minimal parameters and fail with higher level of subdivision, producing vague errors:

MeshLib can subdivide arbitrary meshes without limitations.

6 of 8

Decimation performance

Decimate 1.1M triangles in 100 times till 11K triangles:

VTK: 11 seconds
MeshLib sequential algorithm: 4 seconds
MeshLib parallel algorithm: 1 second

7 of 8

Boolean operations with meshes

VTK library contains an implementation that “relies on the inputs having no co-planar, overlapping cells”, which makes it not practical:�https://vtk.org/doc/nightly/html/classvtkBooleanOperationPolyDataFilter.html
There are third-parties to perform boolean operations in VTK-based software, which proves that VTK-own implementation is very weak:�https://github.com/zippy84/vtkbool
MeshLib on the other hand supports various degenerations in the input meshes, and can even produce a valid result if two meshes coincide in many or all triangles. Also MeshLib implementation can perform boolean operations for moderate-sized meshes in real-time, see MeshLib YouTube channel.

8 of 8

Opening of huge meshes from STL file

In this test we opened a binary STL file of 5Gb containing 105M triangles. Such huge meshes can appear in practice after conversion from volumetric data.

Opening in ParaView 5.10.1 (based on QT+VTK) this operation took at least 10 min (I left the computer after 10 min waiting and returned much later, so probably it took even more).
Opening in MeshInspector (based on MeshLib) this operation took 2 minutes (including rendering).