The frequency characteristics of seismic data play an important role in seismic data processing and interpretation. Many methods have been used to extract a dominant frequency profile. Among these, the instantaneous frequency has been widely used due to its fast calculation and superior vertical resolution. Its incapability of handling noisy data has been rectified by the weighted average method or the generalized Hilbert transform. Several other robust algorithms involve relatively complicated computations, such as sine fitting, wavelet- or spectrum-based filtering, and empirical mode decomposition. Compared to these methods, the local frequency algorithm has the advantage because it handles noisy data or even partially missing data in an elegant and systematic manner. Even though, in cases in which the seismic data energy varies greatly and true amplitude processing is required, the local frequency approach may lose the vertical resolution in weak regions due to the globally controlled smoothing scheme. I have extended the local frequency algorithm with a localized data-adaptive shaping scalar. This data adaptive scheme adds flexibility while preserving all the advantages of the local frequency algorithm. It is capable of removing unrealistic spiky and negative frequencies while keeping the high resolution of the frequency profile. Therefore, it is suitable for records with large energy variations such as true amplitude poststack volumes or prestack shot gathers.