For the given data set and sampled points, our
natural neighbor interpolation routine
outperformed
the built in regularized spline with tension and inverse
distance weighting interpolation routines.
At
a threshold of 10 units, only 5.1% of the pixels were
incorrectly estimated by our interpolation routine
compared with 13.5% and 38.9%
incorrect respectively
for the regularized spline with tension and inverse
distance weighting routines.
At a lower threshold
of 2 units, our routine still dramatically outperformed
the other two methods