Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Sim2Real SAR Image Restoration: Metadata-Driven Models for Joint Despeckling and Sidelobes Reduction

Antoine De Paepe, Pascal Nguyen, Michael Mabelle, Cédric Saleun, Antoine Jouadé, Jean-Christophe Louvigne

TL;DR

The paper tackles joint despeckling and sidelobe reduction for SAR imagery when the forward model is unknown. It introduces a Sim2Real pipeline that trains a DRUNet-based restoration network on MOCEM-generated simulations, including GT $oldsymbol{x}$, pre-transfer SLC $oldsymbol{z}$, observed SLC $oldsymbol{y}$, and acquisition metadata $oldsymbol{m}$, and then applies the model to real SAR data. Metadata is integrated via M-DRUNet and M-SEDRUNet, with losses such as MAE, PL, and EPL guiding optimization; results show metadata-enhanced models achieve higher PSNR on simulated data and improved ENL on real data, indicating effective Sim2Real transfer and controllable restoration. The work highlights the practical value of metadata-aware restoration and motivates future unsupervised diffusion-based approaches to further close the Sim2Real gap in SAR imaging.

Abstract

Synthetic aperture radar (SAR) provides valuable information about the Earth's surface under all weather and illumination conditions. However, the inherent phenomenon of speckle and the presence of sidelobes around bright targets pose challenges for accurate interpretation of SAR imagery. Most existing SAR image restoration methods address despeckling and sidelobes reduction as separate tasks. In this paper, we propose a unified framework that jointly performs both tasks using neural networks (NNs) trained on a realistic SAR simulated dataset generated with MOCEM. Inference can then be performed on real SAR images, demonstrating effective simulation to real (Sim2Real) transferability. Additionally, we incorporate acquisition metadata as auxiliary input to the NNs, demonstrating improved restoration performance.

Sim2Real SAR Image Restoration: Metadata-Driven Models for Joint Despeckling and Sidelobes Reduction

TL;DR

The paper tackles joint despeckling and sidelobe reduction for SAR imagery when the forward model is unknown. It introduces a Sim2Real pipeline that trains a DRUNet-based restoration network on MOCEM-generated simulations, including GT , pre-transfer SLC , observed SLC , and acquisition metadata , and then applies the model to real SAR data. Metadata is integrated via M-DRUNet and M-SEDRUNet, with losses such as MAE, PL, and EPL guiding optimization; results show metadata-enhanced models achieve higher PSNR on simulated data and improved ENL on real data, indicating effective Sim2Real transfer and controllable restoration. The work highlights the practical value of metadata-aware restoration and motivates future unsupervised diffusion-based approaches to further close the Sim2Real gap in SAR imaging.

Abstract

Synthetic aperture radar (SAR) provides valuable information about the Earth's surface under all weather and illumination conditions. However, the inherent phenomenon of speckle and the presence of sidelobes around bright targets pose challenges for accurate interpretation of SAR imagery. Most existing SAR image restoration methods address despeckling and sidelobes reduction as separate tasks. In this paper, we propose a unified framework that jointly performs both tasks using neural networks (NNs) trained on a realistic SAR simulated dataset generated with MOCEM. Inference can then be performed on real SAR images, demonstrating effective simulation to real (Sim2Real) transferability. Additionally, we incorporate acquisition metadata as auxiliary input to the NNs, demonstrating improved restoration performance.
Paper Structure (12 sections, 5 equations, 6 figures, 3 tables)

This paper contains 12 sections, 5 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Problem Formulation
  4. Sim2Real: Learning with Simulations for Real-World Restoration
  5. Input Variations and Custom Loss Functions for Improved Performance
  6. Experiments
  7. Data Preparation
  8. Experimental Settings
  9. Results on Simulated SAR Images
  10. Results on Real SAR Images
  11. Discussion
  12. Conclusion

Figures (6)

  • Figure 1: Images produced with MOCEM.
  • Figure 2: Overview of the architecture of M-SEDRUNet.
  • Figure 3: GT and and restored simulated SAR images.
  • Figure 4: Restored images from real UMBRA SLC.
  • Figure 5: Optical GT (a) alongside CAPELLA SAR image (b) and restored image with SEDRUNet (c).
  • ...and 1 more figures