Least-Squares Diffraction Imaging Using Variational Mode Decomposition

Accurate characterization of subsurface discontinuities is crucial for seismic exploration. However, traditional data-domain diffraction separation and imaging methods encounter challenges when dealing with diffractions and reflections beneath complex overburdens, as the overburdens can alter their distinctive features. We propose a least-squares diffraction imaging (LSDI) method designed to account for differences between diffractions and reflections in both data and image domains. Our approach begins by extending the data-domain diffraction separation method based on variational mode decomposition (VMD) to the image domain. This extension enables separating the diffraction and reflection images based on distinct dip angles. Subsequently, we formulate diffraction imaging as an inverse problem, incorporating data weights to suppress reflections in the common-offset gathers and dynamic model weights to eliminate residual reflectors in the diffraction image. To validate the efficacy of the LSDI approach, we conducted two imaging tests using the SEG/EAGE salt model and field data. The results confirm superiority of the LSDI method over the VMD-based data-domain diffraction imaging method, particularly in terms of suppressing migration artifacts and subsalt reflectors, enhancing image resolution, and recovering weak diffractors.