应用时变子波的盲反射系数反演

由于地震数据的非稳态和全盲特性，实际地震子波通常是未知的且其波形会随传播时间发生变化，故常规基于稳态子波假设的反射系数反演精度将难以保证。为此，利用广义S变换将非稳态地震数据变换到时频域，基于自相关理论逐点提取子波并重新构建时变子波矩阵，而不再是从一道地震记录中仅提取一个时不变子波，将其应用于反演模型以实现非稳态地震资料反射系数的盲反演。模型测试和实际资料处理结果表明，相较于传统时不变子波反演方法，所提方法考虑了地震信号的时变特征，且是数据驱动的，更适用于实际地震资料，能得到更精确的时变子波结果和较高分辨率的反射系数剖面。

A blind deconvolution method based on the time-varying wavelet

Due to the nonstationary and blind seismic data, the actual seismic wavelet is usually unknown and its waveform will change with the propagation time. Therefore, it is difficult to guarantee the accuracy of the inversion results of conventional deconvolution methods based on the stationary wavelet assumption. To this end, nonstationary seismic data were transformed into the time-frequency domain by generalized S-transform. Based on autocorrelation theory, the wavelet was extracted point by point and the time-varying wavelet matrix was reconstructed, instead of extracting only one time-invariant wavelet from one trace of seismic record. Then it was applied to the inversion model for blind deconvolution of nonstationary seismic data. The results of model tests and actual data proces-sing demonstrate that, the proposed data-driven method considers the time-varying characteristics of seismic signals, and it can be more suitable for actual seismic data and get more accurate time-va-rying wavelet results and the reflection coefficient profiles with higher resolution, compared with the traditional time-invariant wavelet inversion method.