Seismic data interpolation using a POCS-guided deep image prior

PreviousNext No AccessSEG Technical Program Expanded Abstracts 2020Seismic data interpolation using a POCS-guided deep image priorAuthors: Min Jun ParkJoseph JenningsBob ClappBiondo BiondiMin Jun ParkStanford UniversitySearch for more papers by this author, Joseph JenningsStanford UniversitySearch for more papers by this author, Bob ClappStanford UniversitySearch for more papers by this author, and Biondo BiondiStanford UniversitySearch for more papers by this authorhttps://doi.org/10.1190/segam2020-3427320.1 SectionsAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail AbstractWe present an algorithm for seismic data interpolation that combines the use of a deep image prior (DIP) and projection onto convex sets (POCS). Deep image priors form part of an optimization problem in which they reparameterize the interpolated data as the output of a convolutional network. While they are able to provide accurate reconstructions of seismic data without the need for any training data, they tend to suffer when large gaps are present in the missing data. We observe significant improvements in the reconstructed data when a POCS regularization term is introduced to the DIP. We demonstrate the improvements of our approach on both synthetic and field data.Presentation Date: Wednesday, October 14, 2020Session Start Time: 8:30 AMPresentation Time: 10:10 AMLocation: 360DPresentation Type: OralKeywords: interpolation, reconstruction, machine learningPermalink: https://doi.org/10.1190/segam2020-3427320.1FiguresReferencesRelatedDetailsCited byImproving sparse representation with deep learning: A workflow for separating strong background interferenceDawei Liu, Wei Wang, Xiaokai Wang, Zhensheng Shi, Mauricio D. Sacchi, and Wenchao Chen27 December 2022 | GEOPHYSICS, Vol. 88, No. 1Self-Attention Deep Image Prior Network for Unsupervised 3-D Seismic Data EnhancementIEEE Transactions on Geoscience and Remote Sensing, Vol. 60Self-Supervised Deep Learning to Reconstruct Seismic Data With Consecutively Missing TracesIEEE Transactions on Geoscience and Remote Sensing, Vol. 60Self-Supervised Learning for Efficient Antialiasing Seismic Data InterpolationIEEE Transactions on Geoscience and Remote Sensing, Vol. 60 SEG Technical Program Expanded Abstracts 2020ISSN (print):1052-3812 ISSN (online):1949-4645Copyright: 2020 Pages: 3887 publication data© 2020 Published in electronic format with permission by the Society of Exploration GeophysicistsPublisher:Society of Exploration Geophysicists HistoryPublished Online: 30 Sep 2020 CITATION INFORMATION Min Jun Park, Joseph Jennings, Bob Clapp, and Biondo Biondi, (2020), "Seismic data interpolation using a POCS-guided deep image prior," SEG Technical Program Expanded Abstracts : 3154-3158. https://doi.org/10.1190/segam2020-3427320.1 Plain-Language Summary Keywordsinterpolationreconstructionmachine learningPDF DownloadLoading ...