Texture analysis and classification of polarimetric SAR images using histogram measures

In polarized synthetic aperture radar (SAR) image classification, texture produces very significant information about land cover. In order to perform the classification in a much better way, a detailed study of histogram measures has been performed. Good accuracies show that histogram measures are used which are best suited for several applications like land use monitoring, surface topology, crustal change, agricultural monitoring, map updating of urban areas, medical field and military system. Histogram measures improve the classification accuracy desirably. PALSAR data, that is unclassified radar images of different polarizations, of Solani River, Roorkee, India and its neighboring regions, is used for classifying the image based on texture features. The textures of images are found on the basis of histogram measures (mean, variance, standard deviation, correlation and skewness). As observed, the area can be characterized into a well-planned dispersed urban area with water bodies and vegetation. The accuracies vary for different classes. For efficient land cover classification, histogram measures have to be chosen suitably. Therefore, the role of various textural histogram measures is analyzed for their discriminative ability for SAR image classification into various land cover types like urban, vegetation and water bodies.