Seismic Data Interpolation Using Streaming Prediction Filter in Time-Space Domain

Abstract

Summary Although the amount of seismic data acquired by wide-azimuth geometry is increasing, it is difficult to achieve regular data distribution in spatial directions, which is limited by surface environment and economic factor. Therefore, interpolation is an important step in seismic data processing. A failed interpolation method may create artifacts, which affect the subsequent processing steps. Now, the mainstream methods of seismic data interpolation are iterative interpolation algorithms, however, the iterative methods always show a slow computational speed that restricts their application in high dimensions. In this paper, we have developed a noniterative four-step strategy to interpolate nonstationary seismic data based on streaming prediction filter (SPF) in time-space domain. Instead of using iterative conjugate-gradient method, we directly calculate the coefficients of nonstationary prediction filter (PF) in the underdetermined equation with local smoothness constraints. On the premise of calculation accuracy, SPF has faster computational speed than iterative methods. Numerical tests using synthetic and field data prove that the proposed method reasonably reconstruct irregularly missing traces with low computational cost even in high-dimensional dataset.