Given a regular grid of squares with sample points at their centres (a pixel-aligned grid), what would be the best way to interpolate to the corners of the squares (a node-aligned grid). I'm aware of all of the generic convolution-kernel methods for interpolating from a grid to arbitrary points, but in the particular case that the interpolation targets are the centre-points of the interpolation source grid, perhaps there are special methods? Or particular convolution kernels? I'd be particularly interested in the case when some of the source values are absent.
[edit after closure]
This question was closed for being "opinion based", but it clearly is not, I ask whether there exist special methods for grid-based interpolation when the target points are the centre points of the interpolation grid squares, appropriate answers could have been "no", or "yes, see [1]". Whoever it was that closed this question should be ashamed of themselves.
