Efficient Methods for Data Smoothing
Larkin B. Scott, Larkin Ridgway Scott · SIAM Journal on Numerical Analysis · 1989
An efficient method for data smoothing via least-squares, polynomial fitting is presented. The method has applications to smoothing experimental data that occur in digital form at uniformly spaced intervals of the independent variable, and it calculates smoothed derivatives of the data with equal efficiency. The work estimate for the method is independent of the length of the smoothing window; thus it avoids explicit convolutions. The storage required is also independent of the window length, except for the actual storage of the original data points.