Diffusion Model Framework for Large-Scale InSAR Phase Unwrapping

A new phase unwrapping framework based on a diffusion model has been developed to address challenges in InSAR processing, particularly for large-scale and complex deformation events. The method, presented in a paper on arXiv (2603.21378), targets scenarios where conventional algorithms fail due to surface-breaking faults and abrupt displacement discontinuities, often caused by shallow earthquake sources. Existing learning-based methods are limited by fixed and small input sizes, which do not match the large-scale and spatially heterogeneous nature of real interferograms. The proposed framework leverages a diffusion model architecture to process large-scale interferograms and handle phase discontinuities, aiming to improve applicability to real-world data. The paper was announced as a replace-cross type and is available at https://arxiv.org/abs/2603.21378.