Statistical Reconstruction of Heterogenous Microstructures Based on 2D Images

Polymer nanocomposites typically possess heterogeneous microstructures that significantly affect structure-property relationships of these material systems. Various microscopic imaging techniques, such as optical microscopy, scanning electron microscopy (SEM), and X-ray microscopy, are essential for characterizing nanocomposite material systems and have provided informative insights of microstructural features. However, microscopic imaging through experiments can be expensive when large amounts of microstructural data are needed. One promising approach to address the imaging limitation and more efficiently generate large microstructural dataset is to statistically reconstruct similar images from a single original input image. A common method used to generate statistically equivalent images is the simulated annealing optimization algorithm. However, due to the high computational cost associated with the stochastic search path used in the simulated annealing algorithm, it can be challenging to reconstruct images with a high degree of agreement. Thus, in this study, a novel and more efficient image reconstruction method was developed by optimizing the simulated annealing algorithm through the manipulation of search path domain and available statistical information. The optimization technique was implemented to reconstruct several example two-dimensional (2D) images to evaluate its capabilities.