We have proposed a new formulation of the seislet transform, a novel
approach to sparse seismic data.
In our approach, we use an RT volume to construct multiscale prediction and
update operators for the seislet transform.
This proposed approach is computationally more efficient and better
preserves geologically meaningful discontinuities than the original
formulation of the seislet transform in reconstructing the raw seismic data.
It also has excellent performance in obtaining the sparse representation of
seismic data.
Synthetic and field data examples show that the proposed new formulation is
efficient and effective in data reconstruction and noise attenuation.