Dually-Constrained Matching Filter for Time-Lapse Seismic Data Consistency Processing

Abstract Time-lapse seismic is widely used to monitor reservoir fluid variations and pore-pressure changes. However, acquisition non-repeatability between baseline and monitor surveys often introduces differences unrelated to reservoir changes. Conventional matching-filter methods can partially reduce these inconsistencies but generally neglect the spatial sparsity of reservoir responses. They are also prone to overfitting and waveform distortion in the presence of non-Gaussian noise, outliers, and signal nonstationarity. To address these limitations, this study proposes a nonconvex matching-filter method based on L1-2 norm regularization. The L1-2 norm is incorporated into a standard matching-filter framework to promote sparsity while controlling filter energy, enabling the method to preserve true reservoir responses and suppress non-reservoir differences. A sliding-window, region-adaptive strategy is further introduced to accommodate signal nonstationarity. The resulting optimization problem is solved using difference-of-convex decomposition combined with the alternating direction method of multipliers (ADMM), ensuring stable convergence. Numerical experiments on both synthetic and offshore field datasets demonstrate that the proposed method significantly improves waveform consistency in non-reservoir regions after matching while effectively preserving true difference signals within reservoir zones. The results further demonstrate that L1-2 norm regularization suppresses background noise without compromising reservoir differences, thereby improving interpretation reliability. These findings indicate that the proposed method is more robust under complex acquisition conditions and provides an effective solution for high-precision time-lapse seismic monitoring.