Accelerating upside-down Rayleigh-Marchenko imaging using self-supervised learning

The upside-down Rayleigh-Marchenko (UD-RM) method has recently been proposed as an effective approach for redatuming and imaging of ocean-bottom seismic data, particularly suitable for acquisition setups with large cable intervals or sparse node geometries. It enables the retrieval of subsurface wavefields and images free from artifacts caused by both internal and surface-related multiples. However, the cost associated with the computation of the required focusing functions is extremely high, especially when targeting large imaging areas. To address this shortcoming, we propose a self-supervised learning approach to accelerate the estimation of the focusing functions. The proposed method is evaluated using a synthetic dataset with both dense and sparse receiver sampling. In both cases, the imaging results are comparable to those obtained with the conventional iterative approach and show a clear improvement over the conventional mirror migration algorithm (when the same input dataset is used). The proposed approach significantly accelerates the UD-RM method, enabling largescale seismic imaging at a fraction of the computational cost of the conventional Marchenko approach.