Spots segmentation in SAR images for remote sensing of environment

Abstract

The paper proposes an algorithm to segment spots in synthetic aperture radar (SAR) images in order to support environmental remote monitoring. This approach consists of isolating dark areas that may have originated from oil pollution, thus achieving the aim of our investigation. The proposed algorithm combines a region growing approach and a multiscale analysis employed by an undecimated wavelet transform to localize dark areas in the sea. The undecimated wavelet applied to SAR images smooths the speckle noise while enhancing edges, thus providing a better result for the proposed segmentation algorithm that is achieved by a modified region growing approach. The minmax scheme is used to provide post processing of the segmented image. The algorithms were tested on real SAR images of oil spills.