<p>Seismic attributes extracted from full frequency seismic data always show weak lateral continuity and unclear distribution when used for identifying faults. Frequency decomposition technology can improve the accuracy of characterizing faults, while the accuracy depends on the time-frequency analysis (TFA) algorithms used. As a widely used TFA method, the generalized S-transform (GST) has the ability to perform multi-resolution analysis, although it still suffers from a problem of insufficient resolution. The time-reassigned synchrosqueezing theory has effectively addressed this problem. Based on this theory, we propose a new time-reassigned synchrosqueezing transform through deriving the group delay operator (GDO) of the GST. In order to achieve the best resolution, we use a parameter matching method to determine the optimal window parameters when calculating the GST spectrum. By performing multiple synchrosqueezing calculations on the obtained spectrum, we finally obtain the time-reassigned multisynchrosqueezing generalized S transform (TMGST). Synthetic signal tests show that TMGST not only exhibits significantly higher resolution than commonly used TFA methods, but also has high flexibility. We use TMGST to extract the frequency decomposition coherence attributes from the Kerry 3D seismic dataset, a publicly available marine seismic survey from New Zealand’s Taranaki Basin, for fault identification. The results show that the coherence anomalies at the fault locations are significantly enhanced and the faults are more clearly characterized by the proposed method in this paper.</p>