Pith. sign in

REVIEW

Poststack Seismic Data Preconditioning via Dynamic Guided Learning

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2502.00887 v1 pith:EI5HEKMJ submitted 2025-02-02 physics.geo-ph

Poststack Seismic Data Preconditioning via Dynamic Guided Learning

classification physics.geo-ph
keywords datanoisetraininglearningpoststackseismicdomaindynamic
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Seismic data preconditioning is essential for subsurface interpretation. It enhances signal quality while attenuating noise, improving the accuracy of geophysical tasks that would otherwise be biased by noise. Although classical poststack seismic data enhancement methods can effectively reduce noise, they rely on predefined statistical distributions, which often fail to capture the complexity of seismic noise. On the other hand, deep learning methods offer an alternative but require large and diverse data sets. Typically, static databases are used for training, introducing domain bias, and limiting adaptability to new noise poststack patterns. This work proposes a novel two-process dynamic training method to overcome these limitations. Our method uses a dynamic database that continuously generates clean and noisy patches during training to guide the learning of a supervised enhancement network. This dynamic-guided learning workflow significantly improves generalization by introducing variability into the training data. In addition, we employ a domain adaptation via a neural style transfer strategy to address the potential challenge of encountering unknown noise domains caused by specific geological configurations. Experimental results demonstrate that our method outperforms state-of-the-art solutions on both synthetic and field data, within and outside the training domain, eliminating reliance on known statistical distributions and enhancing adaptability across diverse data sets of poststack data.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.