Toolbox for interpolating from an uneven grid.

• Interpolation kd-tree implementation for fast nearest-neighbor search in Python. Classes are written for a regularly sampled grid (e.g. Bayer array) or an irregularly sampled grid. The input is a single channel with values at regular or irregular grid points, and the output is multiple channel image with values interpolated from whichever points correspond to that channel (given a lookup table). Can be used as a differentiable PyTorch nn.Module.

• Multi-class Poisson disk sampling on a dense grid. Implemented as described in Wei 2009.

Use the model in model.py to interpolate an image. An example is given there.

Use multiclass_poisson.py for the multi-class Poisson sampler.

@article{wei2010multi,
  title={Multi-class blue noise sampling},
  author={Wei, Li-Yi},
  journal={ACM Transactions on Graphics (TOG)},
  volume={29},
  number={4},
  pages={1--8},
  year={2010},
  publisher={ACM New York, NY, USA}
}