Non-stationary multichannel spectral inversion of seismic data

Spectral inversion, based on the odd-even decomposition principle of reflectivity, used the relationship between seismic data and wavelet amplitude spectrum to establish the inversion equation and achieve resolution-enhancement processing. Compared with deconvolution based on the L₂ norm, the odd and even components of reflectivity using spectral inversion can weaken the tuning effect, identify thin layers, and obtain data with higher resolution. However, most post-stack seismic data are non-stationary, i.e., attenuation of amplitude, phase, and frequency with time exists. We derived a resolution-enhancement algorithm of non-stationary seismic data with quality factor Q based on the short-time Fourier transform. Due to the instability of the spectral inversion algorithm, the lateral continuity of the obtained result is poor. Therefore, we proposed a multichannel spectral inversion algorithm with lateral constraints. The algorithm inherits the high-resolution characteristics of spectral inversion and effectively enhances lateral continuity. Applications to model and field data sets show that the proposed L₂ norm-based non-stationary multichannel spectral inversion method can be effectively applied to the resolution-improvement processing of non-stationary seismic data.