Seismic data interpolation based on U-net with texture loss

Seismic data interpolation is an effective way of recovering missing traces and obtaining enough information for subsequent processing. Unlike traditional methods, deep neural network (DNN)-based methods do not need to make assumptions because they can self-learn the relationship between sampled complete using large training sets with a small computational burden. However, current DNN-based approaches only focus on reducing difference recovered original during training, which helps improve quality reconstructed seismic as whole, while ignoring characteristics local structure. We have developed novel U-net interpolator (SUIT) algorithm based framework DNN in combination texture loss, rather than optimizing reconstruction loss. Texture loss proposed ensure accuracy structural information, calculated by pretrained extraction network. Furthermore, we use trade-off parameter balance error practical technique selecting associated weighting parameter. The feasibility our method assessed via synthetic field examples. Numerical tests show that SUIT robust noisy environments trained reconstruct irregularly or regularly data. Our performed better low-rank matrix fitting method.