Predictive drift compensation of multi-frame STEM via live scan modification

Scanning transmission electron microscopy (STEM) is widely used tool for materials characterisation. However, being a scanned technique, STEM is susceptible to sample, stage or beam drift, manifesting as distortions within images or movement in the field-of-view during multi-frame imaging. Often this is corrected post-acquisition using image registration of multiple frames, but drift reduces the usable area common to all frames. Here we present a method to mitigate sample drift by analysing past frames to predict the sampling-grid points for the immediately future frame. We present this correction across two time-scales and two length-scales. By offsetting the scan-grid framewise we remove long-range drift, and offsetting pixelwise we minimise intra-image warping. Examples are presented for both atomic-resolution imaging and lower-magnification in-situ video capture. The framework is general to raster, serpentine, interlaced and other scan patterns, as well as sequential or scan-rotation series STEM.