Combined multi-branch selective kernel hybrid-pooling skip connection residual network for seismic random noise attenuation

Abstract To improve the generalization ability of single pooling (average or maximum pooling) skip connection residual network (SSN) for seismic random noise attenuation, we present a hybrid-pooling (HSN). In HSN, hybrid consists average and aims to simultaneously capture local global features well, ultimately improving detail recovery capability HSN. further performance denoising propose combined multi-branch selective kernel (CSK) network, which is referred as CHSN. CHSN, CSK three-branch (TSK) our suggested four-branch (FSK), adaptively feature maps high-accuracy effective information recovery. The superior attenuation CHSN demonstrated in both synthetic three- actual two-dimensional data.