Efficient 3-D Seismic Interpolation via Self-Supervised Learning With Dip-Selective Radon-Domain Gating (opens in new tab)

Sub-Nyquist spatial sampling in 3-D seismic acquisition can lead to reflector discontinuities, aliasing artifacts, and acquisition footprints that degrade the subsequent imaging and interpretation. To address these issues, we propose an efficient per-volume self-supervised framework for 3-D seismic interpolation, termed RDG-UNet, which combines long-range coherent-event propagation with dip-selective Radon-domain gating (RDG). The framework is optimized directly on each incomplete volume usin...