Abstract

Summary Multi-dimensional deconvolution (MDD) is a highly desired inversion method that can in principle address multiple challenges in a variety of seismic applications ranging from ocean-bottom data processing, to advanced target-oriented imaging and monitoring. Recent studies show that time-domain MDD, with efficient operator schemes and physics-based constraints, can significantly outperform its more commonly used frequency-domain counterpart in terms of robustness and waveform fidelity when retrieving target responses. Building on this time-domain framework, we propose an alternative MDD scheme based on inversion of point-spread-functions (PSFs), i.e., by recasting the original problem from up/down multidimensional deconvolution into a deblurring problem in the time domain. Due to the symmetry properties of the PSFs in deblurring-based MDD, we show that a relatively simple approach to restricting numerical integration yields inversion results with relatively little loss of accuracy for operators with notably reduced dimensions, which is not the case for the conventional MDD scheme. We illustrate our approach and findings with a numerical example of depth-domain redatuming for target-oriented imaging in a complex subsalt setting. Our approach and results show promise for enabling large-scale time-domain MDD calculations, as well as for bringing more complex MDD applications, such as elastic MDD, into industrial practice.