Abstract:
In this paper we investigate the application of Bayesian segmentation of electron microscopic images based on the Gibbs distribution for visualization of nanostructures. The nature of the Bayesian segmentation is a partition of the image into non overlapping regions, which is the most consistent with the observed image. Quantitative characteristic of this correspondence is a posteriori probability of a particular variant of the partition. The most likely partition is a partition with maximized posteriori probability.