EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Velocity Model Building and Editing With Guided Denoising Diffusion Implicit Models

Francesco Brandolin, Tariq Alkhalifah

Journal of Geophysical Research: Machine Learning and Computation · 2026-10 · 3 (5)

DOI: 10.1029/2026jh001317 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Geophysical Research: Machine Learning and Computation, online erschienen am 16.10.2026. Verfasst von Francesco Brandolin und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 64 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Velocity‐model building is a fundamental component of seismic imaging workflows, yet they remain challenging inverse problems due to limited data coverage, strong nonlinearity, and the need to incorporate heterogeneous sources of information. Such information, like wells, can have a localized influence on the velocity model. In this work, we introduce a unified framework for velocity model editing and velocity model building that combines learned diffusion priors with structurally preconditioned inverse problem formulations. A diffusion model is trained on high‐resolution synthetic velocity models to learn a data‐driven prior, which is then exploited through Denoising Diffusion Implicit Models inversion and guided sampling. For localized velocity model editing, we integrate the diffusion prior with a structurally preconditioned Tikhonov inversion, enabling controlled modification of target regions while preserving the global characteristics of the background model. For full velocity model building, we formulate a well‐matching inverse problem augmented with imaging‐based regularization and solve it using a conventional least‐squares approach, the proposed Denoising Diffusion Implicit Model guided method, and Diffusion Posterior Sampling. Numerical experiments on synthetic benchmarks demonstrate that diffusion‐based approaches consistently recover sharper interfaces and higher‐resolution structures than classical least‐squares inversion. Field data applications on the Viking Graben data set further validate the robustness of the proposed framework for both editing and model building under realistic acquisition conditions.

Zitieren

APA

Brandolin, F., & Alkhalifah, T. (2026). Velocity Model Building and Editing With Guided Denoising Diffusion Implicit Models. Journal of Geophysical Research: Machine Learning and Computation, 3(5). https://doi.org/10.1029/2026jh001317

Vancouver

Brandolin F, Alkhalifah T. Velocity Model Building and Editing With Guided Denoising Diffusion Implicit Models. Journal of Geophysical Research: Machine Learning and Computation. 2026;3(5). doi:10.1029/2026jh001317

BibTeX
@article{Brandolin2026,
  author = {Brandolin, Francesco and Alkhalifah, Tariq},
  title = {Velocity Model Building and Editing With Guided Denoising Diffusion Implicit Models},
  journal = {Journal of Geophysical Research: Machine Learning and Computation},
  year = {2026},
  volume = {3},
  number = {5},
  doi = {10.1029/2026jh001317}
}
Kostenlos starten →