Reflection-waveform inversion based on demigration denoising

Reflection-Waveform Inversion (RWI) is an effective method for constructing high-precision background velocity models by utilizing reflected wave data, overcoming the limitation of conventional Full-Waveform Inversion (FWI) in updating deep background velocities. However, the success of RWI in real data applications is often hindered by low Signal-to-Noise Ratio (SNR) in seismic data. When observed data are contaminated by noise, the predicted data generated through RWI forward modeling fail to adequately match the observed data, leading to erroneous adjoint sources and artifacts in the gradient. This degrades the stability and accuracy of the inversion. To mitigate the impact of noise on RWI, this study introduces migration and demigration as a data preprocessing step. First, the data are mapped to subsurface Offset-Domain Common Image Gathers (ODCIGs) through migration. These gathers are then transformed into the dip-angle domain, where a similarity-based dip filter is applied to suppress noise-induced spurious structures. Finally, demigration is employed to reconstruct the reflected wave signals. The processed data exhibit significantly improved SNR while preserving the traveltime and phase consistency of reflection events in the original data. Tests on both synthetic and field datasets demonstrate that the proposed RWI method based on migrated and demigration data effectively suppresses noise interference during inversion, improving the accuracy of the background velocity model.