A point interpolation algorithm resulting from weighted linear regression

This work presents a novel point interpolation algorithm that is derived from a simple weighted linear regression model. The resulting expression is similar to Inverse Distance Weighting (IDW), which is a widely adopted interpolation algorithm. The novel approach is compared to other methods on synthetic data and also over study cases related to solar radiation, surface elevation, well elevation, and precipitation. Relevant aspects of IDW are preserved while the novel algorithm achieves better results with statistical significance. Artifacts are alleviated in interpolated surfaces generated by the novel approach when compared to the respective surfaces from IDW. The novel method was also revealed, for some cases, as the best alternative among all methods tested in terms of root mean square error. Computational efficiency was shown as competitive or even superior to most of the alternatives under certain conditions. This work is an extended version of our previous conference paper [LNCS 12138, 576 (2020)].