Novel Approach to Interferometric Phase Reconstruction in InSAR Techniques

Researchers at the Hong Kong Polytechnic University have developed a new model for reconstructing interferometric phases in decorrelated regions using a two-stage generative adversarial network (GAN) framework. This breakthrough has the potential to improve topographic retrieval and ground deformation monitoring in interferometric synthetic aperture radar (InSAR) applications. The model's ability to reconnect fragmented phase fringes and reconstruct masked areas has shown promising results in both simulated and real-world experiments.

Key Takeaways:

• The proposed model uses a two-stage GAN framework to reconnect fragmented phase fringes and reconstruct masked areas in interferometric phases.
• The first stage of the model is designed to reconnect fragmented phase fringes, while the second stage focuses on reconstructing the phase in masked regions guided by the reconnected fringes.
• The model was trained on a simulated topographic phase with the SRTAM product and achieved a structural similarity index (SSIM) of 0.9 and a peak signal-to-noise ratio (PSNR) of 30.4.
• The model demonstrated its generalization capabilities in a real-world experiment on a Greater Bay Area (GBA) interferogram, with an average correlation of 0.8 between the predicted and actual phases.
• The proposed approach can effectively preserve phase continuity, reconstruct masked areas, and mitigate the impact of decorrelation in InSAR applications.
• The research has significant implications for improving topographic retrieval and ground deformation monitoring in InSAR applications.

Statistics:

• SSIM (structural similarity index) of 0.9 achieved by the proposed model in simulated experiments.
• PSNR (peak signal-to-noise ratio) of 30.4 achieved by the proposed model in simulated experiments.
• Average correlation of 0.8 between predicted and actual phases in real-world experiments on a GBA interferogram.

Sources:

• A Novel Model for Interferometric Phase Reconstruction Based on Multi-Stage Conditional GANs. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2025,X-G-2025():11-17.
• NewsRx. Hong Kong Polytechnic University Researchers Discuss Research in Information Science (A Novel Model for Interferometric Phase Reconstruction Based on Multi-Stage Conditional GANs). Information Technology Newsweekly. July 29, 2025; p 255.