A novel thresholding method for simultaneous seismic data reconstruction and denoising

摘要

Seismic trace missing together with random noise contamination may seriously influence the performance of some multichannel processing techniques. Simultaneous seismic data reconstruction and denoising is a popular antidote to these two issues since traces missing and random noise both generate random spectra in the frequency domain. However, aliasing in the FK spectrum and random noise residuals are often seen in the reconstructed data with some existing methods. In this paper, we developed a novel thresholding method based on the L-1 norm regularization to improve simultaneous reconstruction and denoising quality. We proposed a modified thresholding method by inducing a weighting parameter to the iterative process. The threshold values are reduced exponentially during iteration to ensure stable outputs. Since the curvelet transform gives a very sparse expression of seismic data, we choose the curvelet transformas the sparse representation. Numerical tests demonstrate that the proposed method achieves both random noise elimination and missing trace reconstruction effectively. (C) 2020 Elsevier B.V. All rights reserved.