Pengliang Yang

RTM using effective boundary saving: A staggered grid GPU implementation

Abstract GPU has become a booming technology in reverse time migration (RTM) to perform the intensive computation. Compared with saving forward modeled wavefield on the disk, RTM via wavefield reconstruction using saved boundaries on device is a more efficient method because computation is much faster than CPU–GPU data transfer. In this paper, we introduce the effective boundary saving strategy in backward reconstruction for RTM. The minimum storage requirement for regular and staggered grid finite difference is determined for perfect reconstruction of the source wavefield. Particularly, we implement RTM using GPU programming, combining staggered finite difference scheme with convolutional perfectly matched layer (CPML) boundary condition. We demonstrate the validity of the proposed approach and CUDA codes with numerical example and imaging of benchmark models.

On analysis-based two-step interpolation methods for randomly sampled seismic data

Interpolating the missing traces of regularly or irregularly sampled seismic record is an exceedingly important issue in the geophysical community. Many modern acquisition and reconstruction methods are designed to exploit the transform domain sparsity of the few randomly recorded but informative seismic data using thresholding techniques. In this paper, to regularize randomly sampled seismic data, we introduce two accelerated, analysis-based two-step interpolation algorithms, the analysis-based FISTA (fast iterative shrinkage-thresholding algorithm) and the FPOCS (fast projection onto convex sets) algorithm from the IST (iterative shrinkage-thresholding) algorithm and the POCS (projection onto convex sets) algorithm. A MATLAB package is developed for the implementation of these thresholding-related interpolation methods. Based on this package, we compare the reconstruction performance of these algorithms, using synthetic and real seismic data. Combined with several thresholding strategies, the accelerated convergence of the proposed methods is also highlighted.

Noise attenuation of the geophysical data using the framelet transform

This paper is on the use of the framelet transformation for noise surpression of geophysical data. The main contribution lies in the application of the transformation for geophysical data. The results of the proposed approach are compared to the use of f-x deconvolution filtering and the curvelet transformation for the same purpose on seismic data. Visually, an improvement in noise reduction is detected.

A Time-Domain Preconditioned Truncated Newton Approach to Visco-acoustic Multiparameter Full Waveform Inversion

A truncated Newton (TRN) method for time-domain full waveform inversion (FWI) in a visco-acoustic medium has been developed based on the 2nd-order adjoint state formulation. Time-domain gradient estimation and Hessian-vector product are managed by recomputing, without numerical instabilities, the incident wavefield at the same time as the adjoint wavefield for mitigating memory issues. Generic algorithm workflow has been proposed to switch between parameterizations, thanks to the chain rule. An efficient preconditioner adapted to the multiparameter configuration is developed to enhance the convergence rate of the inner conjugate gradient iterations. An additional user-defined scaling is introduced in the preconditioner to mitigate the weak sensitivity of specific parameters to waveform variations. The importance of the inverse Hessian for mitigating interparameter trade-off is validated on a toy example. Through a realistic 2D synthetic case based on a North Sea real data application, encouraging numerica...

L1/2-constrained morphological component analysis

In this paper, we introduce the l1/2 constraint into the morphological component analysis (MCA) to solve the problem of cartoon and texture separation, inpainting and super-resolution for images. The newly proposed iterative half thresholding algorithm is recast in an analysis formulation, and can be well embedded in the MCA framework. Other constraint, such as total variation (TV), can also be combined to enhance the restoration of this modified MCA method. The superiority of the proposed l1/2-constrained MCA method is demonstrated by experimental results.

Irregularly sampled seismic data interpolation using iterative half thresholding regularization

The recent progress of compressive sensing (CS) theory shows that perfect reconstruction is possible when the data are sampled randomly and compressed greatly. This revolutionary theory strongly advocates the irregular sampling pattern in the process of seismic data acquisition, leading to a highly reduced acquisition cost. The complete seismic record can be reconstructed from the sparsely few sampled seismic data. In this paper, an analysis version of iterative half thresholding algorithm is introduced to interpolate the seismic data, using the tight frame dictionary. Inspired by the newly proposed spectral compressive sensing theory and the favorable characteristics of tight frame, we employ the redundant Fourier transform to sparsely represent the oscillating seismic data. A double zero-padding strategy in spatial direction of seismic data is suggested to further enhance the reconstruction quality, allowing for the improved interpolation performance and the increased computational cost and storage requirement. The validity of the proposed method is demonstrated by experimental result.