A fast layered alternative to Kriging

Abstract

Empirically gathered scientific data often comes in the form of scattered locations, each with an associated measurement. Visualizing scattered data is challenging because we need to estimate the measured values at many regularly spaced intervals in order to render the data to modern displays. Kriging is a common technique for visualizing scattered data that produces high quality output, but is often too slow for large data sets. In this thesis I present Layered Interpolation, an alternative to Kriging based on the idea of fitting fractal noise functions to scattered data. This technique produces output with quality that is comparable to Kriging, but with greatly reduced running time. Layered Interpolation's speed makes it an ideal choice for rendering large scattered data sets to modern high-resolution displays.