Title: Speeding up of microstructure reconstruction: II. Application to patterns of poly-dispersed islands

Abstract: Recently, a fast reconstruction of a binary labyrinth microstructure by mean of entropic descriptors has been proposed, Piasecki and Olchawa 2012 Modelling Simul. Mater. Sci. Eng. 20 055003. We try to improve the process and make it much faster but for a different class of patterns, i.e. islands of miscellaneous shapes and poly-dispersed in sizes. The hybrid pair of the SD for the spatial inhomogeneity and CS for the statistical complexity is applied yet again. As the second different hybrid pair, we now employ two-point correlation functions, the standard S2 and cluster C2. Since the two pairs appear with weighting factors in the cost function, the present approach is named as the weighted doubly-hybrid, WDH. The competition of the two pairs allows for considering of a wider spectrum of morphological features. Instead of a standard random initial configuration, the synthetic one with same number of islands as of the target is created by a cellular automaton. This is a key point for speeding-up of microstructure reconstruction. Further, the method switches between weak-strong bias modes forced by the current value of interface. This improvement allows for requiring that the value of final and target interface must be the same. The WDH reconstruction terminates when three conditions related to accuracy, interface and number of islands are fulfilled. This method has been verified on digitized images of a thin metallic film and concrete sample cross-section. Our fresh method significantly reduces the entire number of accepted Monte Carlo steps when compared to the standard approach. At the same time, it provides credible shapes and similar areas of islands, keeping their number and the total interface of the target. The reconstructions performed enable us to suggest that the present approach can be used for patterns of islands even with jagged border lines.