Nonstationary seismograms can be mathematically expressed by nonstationary convolution model (NSCM) which integrates equivalent Q-value (Q_e), source wavelet and acoustic impedance (AI) into a single formula. Performing nonstationary AI inversion based on NSCM requires subsurface Q_e field distribution and source wavelet, which are all challenging works. Constant Q_e field and extracted wavelet obtained from the early part of seismograms are commonly used as alternatives, consequently, inaccurate AI inversion results will be easily yielded. To deal with this problem, we decompose nonstationary seismic inversion into three subproblems. Firstly, Q_e curves are estimated at well locations using a novel scheme. Since Gabor transform can induce approximate factorization of NSCM, the time-varying wavelet amplitude spectrum (TVWAS) can be obtained by dividing time-frequency amplitude spectrum (TFAS) of the seismic signal by the TFAS of reflection coefficient. Stable Q_e curves at well locations then can be estimated by the compensation-based Q analysis method whose input data are TVWAS. Secondly, having estimation of Q_e and AI logging information at well locations, source wavelet inversion objective function with wavelet transform domain sparse constraint is proposed based on time domain NSCM. This inverse problem is solved efficiently using projected fast iterative soft-thresholding algorithm (pFISTA). Finally, AI inversion objective function with low frequency constraint and isotropic total variation (ITV) transform domain sparse constraint is presented based on frequency domain NSCM and split Bregman method for minimizing this objective function is proposed. Benchmark model synthetic tests and field data application indicate the effectiveness of our nonstationary AI inversion algorithm by demonstrating the improved accuracy and resolution of the AI inversion results.

Seismic wave will suffer from resolution loss during its propagation in subsurface media, which is caused by the earth filtering effect. Inverse Q-filtering (IQF) is an essential scheme for eliminating earth filtering effect and enhancing temporal resolution. In this letter, IQF is achieved by solving a time-domain inverse problem whose corresponding forward model is established based on the idea of seismic migration, i.e., extrapolating the unfiltered record (the detected signal if there is no earth filtering effect) from surface downward to each time-depth point and then adopting imaging condition to obtain the attenuated record. We express this migration-based forward model as a matrix-vector equation and construct the objective function of inversion-based time-domain IQF (ITIQF). Instead of adopting frequency-domain incomplete amplitude compensation operator for the purpose of antinoise, we propose three transform-domain sparse regularization strategies, including wavelet-based constraint, curvelet-based constraint, and double constraint (a combination of wavelet- and curvelet-based constraints). Solvers of these constrained ITIQF methods are provided according to convex optimization theory. We demonstrate the superior performance of ITIQF scheme via both synthetic and field data examples.

How to effectively remove blending noise is the key to the success of simultaneous source acquisition. The traditional deblending of 3D simultaneous source data is based on the energy difference between blending noise and useful signals in the sparse domain, and remove blending noise independently from each 2D partition of 3D simultaneous source data using sparse inversion algorithm. However, this 3D deblending techniques do not fully utilize a priori information derived from the locations of the spares coefficients from different partitions, especially when highly redundant curvelet domain sparsity is used. In this paper, we propose to use the priori support information to improve the deblending performance of 3D simultaneous source data by using 2D curvelet transform. Firstly, Based on the correlations between the support sets of the significant curvelet coefficients of adjacent partitions, a weighted operator is constructed in the curvelet domain using the prior support sets provided by the previous partition that has removed blending noise. Then incorporate it into the iterative shrinkage thresholding algorithm to delineate the main energy distribution range of the useful signals, and define a sequence of 2D curvelet-based deblending problems that exploit 3D correlations exhibited by seismic wavefields without using the highly redundant 3D curvelet transform. To test the performance of our proposed method, we compared deblending from synthetic and field artificially blended data. These examples demonstrated that our method is superior to the traditional iterative algorithm in terms of deblending performance, and computational memory requirements.

Following the advancement of machine learning-based seismic feature classification techniques for complex reservoirs, the acquisition and analysis of reliable seismic samples involved in seismic facies analysis and network-based inversion have emerged as a current research hotspot in the field of intelligent seismic processing. Many investigations focus on the improvement of model classification algorithms and neural networks. However, creating and collecting labels for massive seismic data are highly time-consuming and laborious, and suffer from sample unreliability and category imbalance in the case of small-sample labels. To address such problems, a multiscale and multilabel consistent principal component analysis-linear discriminant analysis (PCA-LDA) algorithm to learn a robust feature discriminative dictionary for classification is presented. In addition to the automatic use of multilabels from well logs and core analysis, we have associated multiscale with well trajectory locations to enrich sample information and enhance the reliability of the samples during 3D sample acquisition. More specifically, we begin by proposing an approach for the automatic collection of multiscale multilabel 3D poststack seismic samples along the well track. Next, the multilabel sequence in the scan window is fed into the Boyer-Moore majority vote algorithm for sample segmentation, which constructs multilabel hierarchies for each sample. Then to enhance the model training bias due to small-sample label imbalance, we develop a novel label-shuffling balanced strategy, which obtains a complete database by filling random unduplicated augmented training samples (spatial and frequency-domain augmentation operations). Finally, the linear robustness decision-making space of PCA-LDA is obtained using the feature mapping space of PCA, as well as its visual representation. Experimental results on synthetic and field seismic data demonstrate that robust feature extraction with a trustworthy and complete multiscale and multilabel sample database increases classification accuracy.