Riemannian Flow Matching for InSAR phase denoising

Arnaud Breloy (CNAM, CEDRIC laboratory)
June 20, 2025 — 11:00 — "new L2S location (IBM building), Room Hopper (Third floor)" (and Teams)

Abstract

SAR interferometry (InSAR) enables the analysis of topological changes in natural sites or urban areas. However, the use of denoising algorithms is usually required to extract relevant information from interferograms. We will first introduce the application and established multi-temporal InSAR denoising algorithms from the literature, that relate to optimization on the torus on phase-only-vectors. Then we investigate the use of Riemannian flow-matching models for InSAR. Specifically, we consider two use cases: i) the generation of realistic interferograms for data augmentation, ii) the denoising of interferograms. Experiments unconditional generation and conditional generation for denoising of interferograms on Sentinel-1 data illustrate the interest of thes approach.

Bio

Arnaud Breloy is professor at CNAM and the CEDRIC laboratory since 2023. Before that, he was associate professor at Université Paris Nanterre and the LEME labratory (2016-2023). His research focuses on statistics and optimization methods for various applications in machine learning and signal processing.