Texture information-based hybrid methodology for the segmentation of SAR images

Image segmentation is one of the crucial tasks in the postprocessing of synthetic aperture radar (SAR) images. However, SAR images are textural in nature, marked by the textural patterns of widely disparate mean intensity values. This renders conventional multi-resolution techniques inefficient for the segmentation of these images. This article proposes a novel technique of combining both intensity and textural information for effective region classification. To achieve this, two new approaches, called Neighbourhood-based Membership Ambiguity Correction (NMAC) and Dynamic Sliding Window Size Estimation (DSWSE), have been proposed. The results obtained from the two schemes are combined, segregating the image into well-defined regions of distinct textures as well as intensities. Promising results have been obtained over the SAR images of Nordlinger Ries in the Swabian Jura and flood regions near the river Kosi in Bihar, India.