Seismic Multichannel Deconvolution via 2-D K-SVD and MSD-oCSC
Yaojun Wang,Xiayu Gao,Guiqian Zhang,Bangli Zou,Guangmin Hu
DOI: https://doi.org/10.1109/tgrs.2024.3357057
IF: 8.2
2024-02-10
IEEE Transactions on Geoscience and Remote Sensing
Abstract: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 -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.
imaging science & photographic technology,remote sensing,engineering, electrical & electronic,geochemistry & geophysics