Speaker: Giampaolo Ferraioli

Affiliation: Department of Science and Technology, University of Naples Parthenope, Naples, Italy

Report Title: Supervised Deep Learning for SAR Despekling: handling the lack of reference and evaluating the results

Abstract: 

In recent years, the SAR processing domain has been shifting from traditional model-based solutions to deep learning (DL) approaches. While the results achieved so far are impressive, many challenges remain open. A major issue shared by DL-based methods for SAR despeckling is the lack of real ground truth data to serve as a reliable reference. This limitation has two key consequences: (1) supervised training approaches lack the necessary input/reference data, and (2) it is not possible to perform fair and objective comparisons of the obtained results. This tutorial addresses both problems, focusing on the current state-of-the-art solutions as well as emerging perspectives. The goal is to stimulate open discussion on the presented results and highlight promising directions for future research.

Biography: 

Giampaolo Ferraioli was born in Lagonegro, Italy, in 1982. He received the BS and MS degrees and the Ph.D. degree in Telecommunication Engineering. He has been Visiting Scientist at Département TSI of Télécom ParisTech, Paris, France. Currently, he is an Associate Professor with Università degli Studi di Napoli Parthenope. His main research interests deal with Statistical Signal and Image Processing, Radar Systems, Synthetic Aperture Radar, Image Restoration and Deep Learning. He is author of more than 150 papers published on international journals and on international conference proceedings. He won the “IEEE 2009 Best European PhD Thesis in Remote Sensing” prize, sponsored by IEEE Geoscience and Remote Sensing Society. He serves as Associate Editor of IEEE Geoscience and Remote Sensing Letters, as Associate Editor of IEEE Journal on Miniaturization for Air and Space Systems and he is in the Editorial Board of MDPI Remote Sensing.