High-resolution migration image of outside geological structures could be estimated using the measured remote reflection waves. However, they often face contamination from strong direct waves propagating along the borehole, posing challenges for extracting weak reflected waves to obtain the migration image. The paper introduces an approach that operates sequentially in the spatial frequency and Curvelet domains to extract reflected waves. Leveraging VMD, the method iteratively decomposes single-channel spatial domain signals within the COG into high-wavenumber and low-wavenumber IMFs. Subsequently, the low-wavenumber IMF is subtracted from the overall waveform to attenuate direct mode waves. The Curvelet transform is then employed to segregate upgoing and downgoing reflected waves within the filtered Curvelet domain, effectively suppressing direct mode waves while preserving the reflected waves. Through processing synthetic and field data, the efficacy of this approach is validated, demonstrating its potential as a robust extraction technique for enhancing migration image quality.