For classification problems, the law dictates that more proximate pixels are more likely to belong to the same
class. Per-pixel based classifiers, however, do not take spatial context into account, often producing
classified maps with salt and pepper noise (Shekhar et al., 2002)