we present a novel self-supervised learning framework for surface-related multiple suppression. Our method uses a single-pass multi-dimensional convolution to generate synthetic surface multiples and incorporates them into a two-stage training strategy (warm-up and iterative data refinement) that does not require labeled data or detailed subsurface models. We demonstrate its effectiveness on both synthetic and real marine datasets.
GitHub: https://github.com/DeepWave-KAUST/SSL-Multiples-Attenuation-pub