The effectiveness of full-waveform inversion (FWI) and imaging in land seismic data is highly dependent on the data quality. However, land seismic datasets are often characterized by a low signal-to-noise ratio (S/N) due to factors such as strong near-surface velocity contrasts, elastic effects, and subsurface heterogeneity, which significantly hinder the accuracy of both inversion and imaging processes. In this study, we propose a sparsity-promoting weighted Radon preconditioning method to enhance the quality of early-arriving seismic signals for FWI. We then employ demigration using an extended-imaging condition to improve the reflection quality for land seismic imaging. The proposed method combines a scanning-based approach for isolating early arrivals with a semblance-weighted sparse Radon transform. The scanning technique automatically guides early arrival identification based on maximum stacking energy and linear velocity assumption. The semblance-weighted sparse Radon transform is then applied to enhance the coherence of diving and post-critical early arrivals used in FWI. Numerical tests on both synthetic and field datasets demonstrate that our method significantly improves the coherency and S/N of early arrivals, as well as the quality of low-frequency components. These improvements in early arrivals enable FWI to generate a high-fidelity near-surface velocity model with enhanced resolution for land seismic data. Furthermore, demigration using extended-imaging condition also greatly enhances the reflection coherency and suppresses linear noise interfering with reflections, leading focused seismic images of land data.

Full waveform inversion (FWI) is a notable technique that provides high-resolution physical parameters of subsurface media. Although FWI is frequently employed to recover velocity models for relatively weak parameter perturbations, its effectiveness is limited by the lack of low-frequency information in the presence of strong parameter perturbations. To address this limitation, we propose to utilize direct envelope inversion (DEI), which highlights the low-frequency information contained within the seismic envelope data to successfully construct velocity models for strong parameter perturbations. However, conventional DEI requires envelope computation for the source wavefield, which limits the application of memory cost-reduction methods and significantly increases the computational time needed for the envelope. To mitigate these computational challenges, we introduce excitation amplitude (ExA) as a means to reduce the computational cost associated with DEI. By utilizing only the most energetic amplitude and its arrival time at each grid point of the direct envelope virtual source field, this method can reduce the computational time and memory requirement while maintaining the accuracy of the DEI. In the numerical examples, we demonstrate that the proposed method overcomes the computational cost limitations of conventional DEI. Additionally, applying our method to field data acquired in the Arctic helped reconstruct strong scattering models for the subsea permafrost.

Least-squares reverse time migration (LSRTM) is linearized inversion based on Born approximation, which commonly seeks to high-quality migration result by least-squares sense. Contrary to time domain LSRTM which performs the wavefield simulation as much as several times of the number of shots, frequency-domain LSRTM (F-LSRTM) has advantage to efficiently deal with multiple shot records. Furthermore, if Green’s function can be saved on memory storage, the full wavefield simulation is implemented only once at the first iteration during entire LSRTM iterations. However, huge memory storage may be required to save the Green’s function for large dataset and model size. To alleviate this computational issue, we propose an efficient F-LSRTM scheme using singular value decomposition (SVD). In our scheme, Green’s function can be saved efficiently as two unitary matrices and one singular value vector with a few number of dominant singular values. Because the number of dominant singular values, called as the optimal rank, is much smaller than the minimum value in each dimension size of Green’s function, the proposed method can make it possible to save the Green’s function into the computing memory with keeping the accuracy. After demonstrating the feasibility of reducing the rank of Green’s function, we examine our proposed F-LSRTM scheme and comparative F-LSRTM schemes (F-LSRTM using adjoint-state method and saved full Green’s function, respectively) using the simple layered and modified marmousi-2 model. Numerical tests indicate that our proposed F-LSRTM scheme can generate migration results as accurate as comparative F-LSRTM schemes with less memory usages and computational cost.

Optical diffraction tomography measures the three-dimensional refractive index map of a specimen and visualizes biochemical phenomena at the nanoscale in a non-destructive manner. One major drawback of optical diffraction tomography is poor axial resolution due to limited access to the three-dimensional optical transfer function. This missing cone problem has been addressed through regularization algorithms that use a priori information, such as non-negativity and sample smoothness. However, the iterative nature of these algorithms and their parameter dependency make real-time visualization impossible. In this article, we propose and experimentally demonstrate a deep neural network, which we term DeepRegularizer, that rapidly improves the resolution of a three-dimensional refractive index map. Trained with pairs of datasets (a raw refractive index tomogram and a resolution-enhanced refractive index tomogram via the iterative total variation algorithm), the three-dimensional U-net-based convolutional neural network learns a transformation between the two tomogram domains. The feasibility and generalizability of our network are demonstrated using bacterial cells and a human leukaemic cell line, and by validating the model across different samples. DeepRegularizer offers more than an order of magnitude faster regularization performance compared to the conventional iterative method. We envision that the proposed data-driven approach can bypass the high time complexity of various image reconstructions in other imaging modalities.