Algorithms for Crystallography in the Scanning Electron Microscope

 This thesis explores new approaches to enhance the efficiency and accuracy of materials characterization in the scanning electron microscope (SEM) with particular emphasis on electron backscatter diffraction (EBSD). The work addresses two challenges in the field: accelerating data acquisition and improving the handling of orientation data acquired from polycrystals. Firstly, we introduce an unsupervised dynamic sampling approach based on a nearest neighbor heuristic for scanned microscopy modalities. Secondly, we present a variation on dictionary-based orientation indexing for EBSD using principal component analysis (PCA) and numerical quantization which accelerates and improves the noise robustness of the method. Thirdly, we cover two potential tools for multimodal image registration in the SEM using approaches rooted in information theory. Lastly, we develop decision tree algorithms for the efficient reduction of crystallographic orientations to the fundamental zone of their respective Laue classes. By addressing key algorithmic obstacles in SEM-based microstructure characterization, this thesis aims to alleviate some of the challenges of handling such materials science data in this era of big data.