The class statistics and Euclidean measure are both used to determine the neighborhood of a pixel to keep classes separability and preserve data distribution.