Incorporating Physical Constraints and Regularization in Min-cut/Max-flow Graph Partitioning for Segmentation and Clustering in Materials ImagingMS11

Engineering materials are heterogeneous mixtures at the microscale. Often the properties of the material have a strong dependence on features which have been obscured by phase transformations. Reconstruction of the prior structure is ill-posed, as the forward transformation is one-to-many. Here we present a graph partitioning approach to incorporate physically motivated priors in stochastic image reconstruction. The approach will be demonstrated on recovery of pre-transformation microstructures as well as segmentation and clustering in materials microscopy.

This presentation is part of Minisymposium “MS11 - Computational Imaging for Micro- and Nano-structures in Materials Science (2 parts)
organized by: Brendt Wohlberg (Los Alamos National Laboratory) , Jeff Simmons (Air Force Research Laboratory) .

Stephen Niezgoda (The Ohio State University)
Alexander Brust (The Ohio State University)
Eric Payton (Air Force Research Laboratory)
bayesian methods, electron diffraction, electron microscopy, image reconstruction, image segmentation, inverse problems, statistical inverse estimation methods