Compressed Sensing for Fast Electron Microscopy

Scanning electron microscopes (SEMs) are used in neuroscience and materials science to image square centimeters of sample area at nanometer scales. Since imaging rates are in large part SNR-limited, imaging time is proportional to the number of measurements taken of each sample; in a traditional SEM, large collections can lead to weeks of aroundthe- clock imaging time. We previously reported a single-beam sparse sampling approach that we have demonstrated on an operational SEM for collecting "smooth" images. In this paper, we analyze how measurements from a hypothetical multi-beam system would compare to the single-beam approach in a compressed sensing framework. To that end, multi-beam measurements are synthesized on a single-beam SEM, and fidelity of reconstructed images are compared to the previously demonstrated approach. Since taking fewer measurements comes at the cost of reduced SNR, image fidelity as a function of undersampling ratio is reported.