Efficient Magnitude-Domain Plug-and-Play Regularization Method for SAR Imaging SAR Görüntüleme için Büyüklük Uzayinda Etkin Tak-Çaliştir Düzenlileştirme Yöntemi
34th Signal Processing and Communications Applications Conference, SIU 2026, İstanbul, Turkey, 7 - 10 July 2026, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/siu71813.2026.11637077
- City: İstanbul
- Country: Turkey
- Keywords: deep priors, image reconstruction, inverse problems, plug-and-play regularization, Synthetic aperture radar
- Middle East Technical University Affiliated: Yes
Abstract
Radar imaging seeks to reconstruct the reflectivity distribution of a scene using measurements of backscattered electromagnetic waves. In this study, an efficient plug-and-play reconstruction approach with deep priors is adapted to synthetic aperture radar (SAR) imaging. For this purpose, the measurement model is first presented. For the solution of the inverse problem corresponding to the image reconstruction problem, regularization is enforced in the magnitude domain due to the random nature of the phase. In the adapted ADMM-based reconstruction approach, for the proximal operator, a closed-form efficient solution is used for the phase update step, while a U-Net-based denoiser is used for the magnitude component. Based on the training and experiments performed on Sentinel-1 SAR images, it is observed that the developed method provides higher image quality than classical FFT-based reconstruction at different data availability and noise levels.