L-BFGS BASED LEAST-SQUARES REVERSE TIME MIGRATION FOR REFLECTIVITY IMAGING

Resumo

Assuming kinematic accuracy on the migration velocity model, seismic least-squares migration seeks to
overcome the drawbacks of traditional migration algorithms by approximating the inverse imaging operator within
a linear inversion framework. In this paper, we combine reverse-time migration and the L-BFGS optimization
method to iteratively reconstruct the true subsurface reflectivity model from seismic reflection measurements. In
each iteration, the initial model is first demigrated using the Born approximation of the acoustic wave equation to
generate the predicted seismic data. Then, the reflectivity is updated by minimizing the least-squares misfit function
in the data domain. Furthermore, to speed up the convergence of the inversion algorithm, we preconditioned the
gradient by the illumination compensation operator. Numerical examples demonstrate that, even in the presence
of salt bodies, the inverse operator of the imaging problem can be used to obtain improved migrated sections with
reduced artifacts, better resolution, and reflectors with more balanced amplitudes.