Objective and Methods Normal moveout (NMO) correction is a core step in seismic data processing. However, the conventional NMO correction method tends to cause wavelet stretching when processing wide-angle or large-offset seismic data, leading to degraded stacking quality and distorted amplitude variation with offset (AVO) attributes. Although the matching pursuit (MP)-NMO correction method can mitigate the stretching effect, it will cause lateral discontinuity due to overlapping seismic events. Therefore, this study developed an improved MP-NMO correction method based on multiparameter fitting. Specifically, during MP decomposition, cubic polynomial fitting was applied to the amplitude, frequencies, and phases of wavelets from the same reflection interface across different offsets, thus eliminating local anomalies. With the second-order time-offset curve as the initial value, the optimal arrival time of reflected waves was searched within the dynamic window by combining the adaptive time scanning strategy, thus avoiding the complexity of solving higher-order time-offset equations. Additionally, the multiscale scanning of the complex Morlet wavelet was introduced to enhance noise resistance and adaptability to frequency bands.

Results Theoretical model-based experiments demonstrate that the improved MP-NMO correction method effectively mitigated wavelet stretching under large offsets, yielding amplitude and frequency errors of less than 5%, which decreased by more than 90% compared to those of the conventional MP-NMO correction method (up to 40%). Furthermore, the improved method exhibited enhanced noise resistance and achieved convergence within seven iterations. When applied to actual data, the improved method enhanced the continuity of seismic events in large-offset gathers of shallow structures and the resolution in stacked seismic sections, yielding clearer images of events at critical intervals between 0.8 s and 1.7 s.

Conclusions The results of this study provide a technical method for the high-fidelity processing of seismic data from structurally complex areas. This method is especially applicable to the analysis and anisotropic inversion of AVO/amplitude versus frequency (AVF) attributes.