ABSTRACT High‐accuracy data reconstruction is a crucial step in enhancing the performance of subsequent seismic processing methods. In three‐dimensional (3D) seismic data reconstruction, conventional methods typically reconstruct each slice independently in the time domain. However, seismic data in the time domain exhibit relatively low sparsity. Moreover, the correlations between adjacent slices are neglected, thus limiting the reconstruction quality. To address this issue, we propose a novel 3D seismic data reconstruction method in the frequency domain. The method first applies a Fourier transform to convert the incomplete seismic dataset from the time domain into the frequency domain. Then, utilizing the projection onto convex sets (POCS) algorithm and the multiscale curvelet transform, and leveraging the substantial overlap in the curvelet coefficient supports of adjacent frequency slices, we construct a frequency domain–weighted operator using the prior support information from a previously reconstructed slice. This operator delineates the effective signal energy distribution and provides prior support set information for reconstructing the next frequency slice. Finally, an inverse Fourier transform is applied to obtain the reconstructed 3D seismic data in the time domain. Comparative analyses of synthetic and field data demonstrate that the proposed method significantly improves reconstruction performance, outperforming conventional time domain seismic data reconstruction methods.
Zikun et al. (Wed,) studied this question.