Application and further development of advanced image processing algorithms for automated analysis of serial section image data

Several automated algorithms are presented for the segmentation of features of interest from microstructure images acquired with modern high-throughput electron microscopes. Specifically, the maximization of posterior marginals (MPM) segmentation technique, originally developed for computer vision applications, is applied towards automated segmentation of microstructural images from Ti- and Ni-alloy systems. The MPM technique classifies image pixels according to the most probable class to which they can belong. Three derivatives of the MPM algorithm are introduced and assessed: expectation maximization MPM (EM/MPM), EM/MPM with simulated annealing (EM/MPM/SA) and vector EM/MPM/SA. Example applications of all three approaches are given. The EM/MPM model allows for automated segmentation of α laths in a Ti-6242 sample and primary γ' in an IN100 superalloy, but has difficulty accurately locating the boundaries between regions. The EM/MPM/SA algorithm involves a gradual increase in the interface capillarity during segmentation and allows for pixel accuracy determination of boundaries between phases. The vector EM/MPM/SA method is capable of simultaneously segmenting a series of images acquired with differing imaging conditions. The limitations of the algorithms are discussed as well as potential future modifications.