Segmentation of regions of interest on locally homogeneous scenes

Abstract

The possibilities of the segmentation of regions of interest on locally homogeneous scenes using some known methods are studied. The segmentation is considered a problem of classifying pixels of the region of interest into two classes and the segmentation quality is measured by the probability of misclassification. The results of the segmentation for scenes with different correlation radii and complexities, which is measured as the signal-to-noise ratio, are discussed. © 2012 Pleiades Publishing, Ltd.