Abstract:Seismic data often undergo severe noise due to environmental factors, which seriously affect subsequent applications. Traditional hand-crafted denoisers such as filters and regularizations utilize interpretable domain knowledge to design generalizable denoising techniques, while their representation capacities may be inferior to deep learning denoisers, which can learn complex and representative denoising mappings from abundant training pairs. However, due to the scarcity of high-quality training pairs, deep learning denoisers may sustain some generalization issues over various scenarios. In this work, we propose a self-supervised method that combines the capacities of the deep denoiser and the generalization abilities of the hand-crafted regularization for seismic data noise attenuation. Specifically, we leverage the Self2Self (S2S) learning framework with a trace-wise masking strategy for seismic data denoising by solely using the observed noisy data. Parallelly, we suggest the weighted total variation (WTV) to further capture the horizontal local smooth structure of seismic data. Our method, dubbed as S2S-WTV, enjoys both high representation abilities brought from the self-supervised deep network and good generalization abilities of the hand-crafted WTV regularizer and the self-supervised nature. Therefore, our method can effectively remove noise and preserve the fine details of the seismic signal. To tackle the S2S-WTV optimization model, we introduce an alternating direction multiplier method (ADMM)-based algorithm. Extensive experiments on synthetic and field noisy seismic data demonstrate the effectiveness of our method as compared with the state-of-the-art traditional and deep learning-based seismic data denoising methods. Code is available at https://github.com/XuZitai/S2S-WTV.

S2S-WTV: Seismic Data Noise Attenuation Using Weighted Total Variation Regularized Self-Supervised Learning

Residual Learning of Deep Convolutional Neural Network for Seismic Random Noise Attenuation

Deep Nonlocal Regularizer: A Self-Supervised Learning Method for 3-D Seismic Denoising

A Denoising Method for Seismic Data Based on SVD and Deep Learning

Bilateral Spectrum Weighted Total Variation for Noisy-Image Super-Resolution and Image Denoising

Self-Supervised Attenuation Method Based on Similarity Comparisons for 3D Seismic Random Noise

Self-adaptive denoising net: Self-supervised learning for seismic migration artifacts and random noise attenuation

NS2NS: Self-Learning for Seismic Image Denoising

Seismic Data Denoising Based on Tensor Decomposition With Total Variation

Unpaired training: Optimize the seismic data denoising model without paired training data

Random Noise Attenuation Based on Support Vector Regression and Adaptive Wiener Filtering

Self-Supervised Seismic Random Noise Attenuation With Spatial Attention From a Single Section

Regularized deep learning for unsupervised random noise attenuation in poststack seismic data

Hyperspectral image denoising via weighted double sparsity total variation and low-rank representation

An effective self-supervised learning method for various seismic noise attenuation

Hyperspectral image denoising using constraint smooth rank approximation and weighted enhance 3DTV

A two-stage seismic data denoising network based on deep learning

Unsupervised Deep Learning for Random Noise Attenuation of Seismic Data

Self-Supervised Seismic Swell Noise Suppression From Noisy Seismic Data

An Adaptive High-Dimensional Progressive Denoising Method for Seismic Weak Signal Enhancement

High-Order Directional Total Variation for Seismic Noise Attenuation