Minimizing image-processing artifacts in scanning tunneling microscopy using linear-regression fitting

We present a method for removing noise from scanning tunneling microscopy images based on least-squares fitting of spatial data. Modeling the known structure of the surface, including isolated features and surface steps, allows for effective discrimination of signal from noise and produces minimal processing artifacts, even for very noisy images. This approach is effective for removing external noise due to vibrational or acoustic interference, and can also be applied to correct for tip-related height jumps as well as to flatten images warped by thermal effects or nonlinearity of the microscope scanner.