Seismic data interpolation beyond continuous missing data using high-order streaming prediction filter

Wu Geng,Liu Cai,Liu DianMi,Liu Yang,Zheng ZhiSheng

DOI: https://doi.org/10.6038/cjg2022Q0458

2023-01-01

Abstract:Due to the limitations of geometry implementation and economic factors, the acquired seismic data are always irregularly distributed along the spatial directions and often have a large range of continuous missing traces. Many subsequent data processing methods (e. g., multiple wave suppression and wave equation migration, etc.) require spatially regularly distributed data. Interpolation techniques are effective approaches to solve the missing seismic data problem, but traditional data interpolation methods occasionally fail when performing continuous missing data reconstruction, especially when dealing with non-stationary seismic events with low accuracy. Meanwhile, most interpolation methods require iterative computation, which is less efficient when dealing with high-dimensional large-scale data. To address the fast interpolation problem of continuous missing data, this paper proposes a noniterative interpolation method using high-order streaming prediction filter in the space-time domain, which improves the filter estimation accuracy of continuous missing data by using high-order constraints. The proposed method improves the stability of local constraints and the invalid situation caused by low-order streaming computation due to the inability to continuously update the filter coefficients. At the same time, the spatially noncausal filter and snaky interpolation path are designed to reduce the computation error of prediction filters from the large range of continuous missing data and data boundaries. Therefore, this method can effectively handle the reconstruction of continuous missing data including the missing near-offset case. Compared to the industry standard Fourier Projection onto Convex Sets (POCS) method, the proposed interpolation method based on high -order streaming prediction filter has a better reconstruction effect on the high-dimensional continuous missing seismic data, which shows a better balance in both interpolation accuracy and computational efficiency according to the synthetic model and field data tests.