Plan for applying gaussian filter to adaptive grids

[milos7250]

Applying the Gaussian Filter

• Pre-process the data to have the same spatial resolution (using interpolation of some kind)
• Create method that only loads necessary subsets of data from disk to memory to apply the filter to the whole data. #9
  • Balance memory usage with disk I/O

Creating a Triangular Mesh from the Smoothened Data

• Create a method to generate triangular mesh from single subset of data in spherical coordinates
  • Consider using other meshing algorithms (inspiration) #10
• Create a method to merge duplicate vertices in a triangular mesh
  • Add functionality for normals and values
  • Investigate and optimize memory efficiency

[milos7250]

The merging of duplicate vertices is not necessary. In mayavi, visually it makes no difference if the duplicate points are removed or not. In Unity, the duplicate vertices are removed automatically on import. Unnecessary faces can be removed by supplying allow_degenerate=False to the skimage marching cube algorithm.

[milos7250]

The method to load and merge different grids is here. As of now, only grids with the same spatial resolution are supported. The grids must also overlap at the gridpoints precisely. This function will not work if e.g. one grid covers points with r coordinates 0.5, 1.5, 2.5 and the other covers 1, 2, 3, as overlapping them would produce a grid that is not uniform in spatial resolution anymore. Getting rid of this restriction is tracked in #11.

As so far there hasn't been problems with running out of memory when applying the gaussian filter, #9 is kept as an enhancement proposition.