search

ddemidov / mba

Scattered data interpolation with multilevel B-Splines
MIT License
73 stars 23 forks source link

Normalization #15

Open untereiner opened 3 years ago

untereiner commented 3 years ago

I open a new issue to not mix the two questions

It is very strange because my example points aren't working until I normalize them. The operator interp(mba::point<2>{u, v}) with large values of u and v returns 0.

ddemidov commented 3 years ago

Are large values of u and v still inside your domain (defined with cmin and cmax)? If not then these are outside of interpolation grid and decrease rapidly.

untereiner commented 3 years ago

My dataset is composed of 3D points. The cmin, cmax are the min/max of the x,y bounding box and the values vector are the z coordinates of my points. So yes they should

ddemidov commented 3 years ago

Another problem could be the initial grid is too sparse, so the data points are approximated exactly, and the rest of the grid has no overlap with the input data.

terewenko commented 3 years ago

There was the same problem, it turned out to be solved by reducing the grid (mba:: index<2> grid).