Seismic Multichannel Deconvolution via 2-D K-SVD and MSD-oCSC

The deconvolution method is crucial for enhancing seismic resolution. Traditional multichannel schemes incorporate lateral constraints to enhance data continuity and integrity. However, a predominant challenge is that many existing methods assume that the estimated model parameters have certain feat...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE transactions on geoscience and remote sensing Ročník 62; s. 1 - 13
Hlavní autoři: Wang, Yaojun, Gao, Xiayu, Zhang, Guiqian, Zou, Bangli, Hu, Guangmin
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York IEEE 2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Témata:
2-D K-means singular value decomposition (K-SVD)
Algorithms
Coding
Components
Convolution
Convolutional codes
convolutional sparse coding (CSC)
Decomposition
Deconvolution
Dictionaries
Encoding
Feature maps
Frequencies
Glossaries
Image coding
Iterative methods
Linear programming
multichannel
Objective function
Reflectivity
Regularization
Segments
Seismic analysis
Seismic data
seismic deconvolution
Seismic profiles
Seismology
Singular value decomposition
ISSN:0196-2892, 1558-0644
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The deconvolution method is crucial for enhancing seismic resolution. Traditional multichannel schemes incorporate lateral constraints to enhance data continuity and integrity. However, a predominant challenge is that many existing methods assume that the estimated model parameters have certain features, which are always inconsistent with the true situation. In this article, we present a novel data-driven multichannel seismic deconvolution method based on convolutional sparse coding (CSC) and 2-D <inline-formula> <tex-math notation="LaTeX">K </tex-math></inline-formula>-means singular value decomposition (K-SVD). We partition the seismic profile into low- and high-frequency segments, addressing CSC's limitation in representing low-frequency components effectively. We leverage 2-D K-SVD for extracting lateral features in low-frequency components and apply CSC to the high-frequency segments. Distinctively, our method accounts for the lateral characteristics within both frequency bands during decomposition, a notable advancement over conventional filtering-based techniques. It ensures the preserved convolutional relationship in frequency separation. We enhance the conventional 1-D K-SVD with 2-D sample patches, evolving it into a more adept 2-D K-SVD dictionary for low-frequency components. We apply this to multichannel deconvolution regularization, enabling low-frequency seismic data deconvolution. To address the challenge of information fragmentation in high-frequency components caused by block-based sparse coding, the orthogonal constrains are added into the feature map instead of the image, making the convolutional dictionary learning more precise than the conventional CSC algorithm. This refinement is integral to optimizing the high-frequency deconvolution objective function. In the final step, iterative processing of both frequency components yields the refined multichannel seismic deconvolution via orthogonal convolution sparse coding (MSD-oCSC). Our method's efficacy is corroborated through rigorous model tests and real data applications.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0196-2892
1558-0644
DOI:10.1109/TGRS.2024.3357057