Multichannel Prestack Attenuation Compensation Based on Low-Dimensional Manifold Model

The quality of seismic signal is seriously affected by attenuation and random noise. For prestack seismic signal, the Earth filtering effect reduces the vertical seismic resolution, as well as distorts the amplitude variation with angle (AVA). Furthermore, since prestack seismic signal generally has lower signal-to-noise ratio (SNR), noise is more likely to be amplified when eliminating this absorption attenuation effect. Therefore, we propose a robust multichannel prestack compensation method to simultaneously remove noise and compensate attenuation under the inversion framework. Instead of using incomplete compensation operator to process seismic signal, we construct a multichannel inversion problem by connecting attenuated and un-attenuated seismic signal via convolution and time-domain attenuation function. Moreover, considering the inherent low-dimensional structure with patch manifolds of seismic signal, the dimension of the patch manifolds is used as a regularization term to obtain stable compensation results. This method demonstrates strong ability of random noise suppression and effective signal preservation during the compensation process. Numerical experiment results on synthetic and field data show that compared with the prestack inverse Q filtering (IQF), this method provides better compensation results with higher SNR and accuracy.