Exemplar-Based Inpainting Based on Dictionary Learning for Sparse Scanning Electron Microscopy

Abstract

High-throughput scanning electron microscopy (SEM) aims to reduce dose for sensitive specimens as well as reducing acquisition times to be able to acquire large volumes in a meaningful time. Sparse sampling is one key to make such acquisitions possible. We propose a new reconstruction technique for such sparsely sampled SEM data, which is based on exemplar-based inpainting known from image processing.