Sparse unscaled time-frequency transform and its application on seismic attenuation delineation

Seismic attenuation can be qualitatively delineated by the difference between low- and high-frequency components before and after attenuation. Time-frequency (TF) transform is a useful tool that is commonly used in the characterization of frequency component changes caused by seismic attenuation, in which the representative S-transform (ST) often is used for solving this issue. However, ST has a fixed TF localization and a significant frequency shift issue, leading to low resolution and accuracy of the results when used for qualitative attenuation estimation. To solve these issues, this study introduces an unscaled generalized ST (UGST) based on the unscaled ST to avoid the dominant frequency shift and achieve flexible adjustment of the TF localization. To further improve the resolution of TF, the solution of UGST is regarded as a convex optimization process with L1-L2-norm joint constraints, and the sparse UGST (SUGST) is solved by using the Bregman iterative algorithm. Synthetic experiments indicate that SUGST has high TF resolution and dominant frequency fidelity. Qualitative attenuation delineations of synthetic and field data are consistent with the synthetic model and reservoir interpretation. Its resolution and accuracy also are better than contrast methods, which further demonstrate the validity and advantage of the SUGST-based workflow in seismic attenuation delineation and reservoir interpretation.