Recent Advances in SAR Image Despeckling and Change Detection Using Deep Learning Approaches

Deep learning is a machine learning technique that has significantly improved results in many areas such as computer vision, speech recognition, machine translation, and biomedical imaging analysis and understanding. Recently, in the field of synthetic aperture radar (SAR) images analysis, deep learning approaches becomes a powerful tool making information extraction from SAR images very accurate and give good interpretations related to environment and earth surface observation. This work examines the recent development of scientific productions on the applications of deep learning approaches in the SAR imaging field. These applications concern the speckle noise reduction from SAR images and the algorithms-based method for change detection and classification of these remote sensing images. Moreover, an analysis of scientific production in this field is discussed by exploiting IEEE Xplore digital library and SCOPUS database. An exploitation of convolutional neural network (CNN) and transfer learning with residual network-based method for despeckling and flooded zone detection is presented in the end of the paper.