Time-reassigned synchrosqueezing tunable-factor Gabor wavelet transform for thin-layer analysis of seismic data

Abstract

Reducing uncertainty in thin-layer geological interpretation requires detailed characterization of the spatial distribution of thin layers. Time-frequency (TF) analysis (TFA) reveals the energy distribution and variation of seismic signals across time and frequency, facilitating the extraction of thin-layer information. However, conventional TFA methods suffer from interlayer interference, making accurate spatial characterization of thin-layer interfaces challenging. Therefore, a method termed NMF-TSTFGWT is proposed, which incorporates non-negative matrix factorization (NMF) into the time-reassigned synchrosqueezing tunable-factor Gabor wavelet transform (TSTFGWT) to enhance the TF representation (TFR). Within TSTFGWT, the tunable-factor Gabor wavelet provides locally adaptive temporal resolution. The resulting tunable-factor Gabor wavelet coefficients are then reassigned along the temporal dimension using synchrosqueezing, thereby enhancing temporal resolution. Furthermore, the magnitude of the TSTFGWT coefficients is computed to construct the TFR, and NMF is employed to reduce redundancy in the spectral features. This low-rank decomposition captures the TFR's dominant features, suppresses spectral interference, and improves TF resolution. Experiments on synthetic and field seismic data show that the proposed NMF-TSTFGWT improves energy concentration in the TFR and yields more accurate thin-layer characterization. Moreover, derived thin-layer interfaces are consistent with well-log interpretations, effectively delineating thin-layer interfaces and improving the reliability of geological interpretations.