To suppress the strong noise in seismic data with wide range of amplitudes, commonly used methods often yield unsatisfactory denoising results owing to inappropriate thresholds and require parametric testing as well as iterations to achieve the anticipated results. To overcome these problems, a data-driven strong amplitude suppression method based on the decibel criterion in the wavelet domain （ISANA） is proposed. The method determines the denoising threshold based on the decibel criterion and statistically analyzes the amplitude index rather than the abnormally high amplitudes. The method distinguishes the frequency band distributions of the valid signals in the time-frequency domain based on the wavelet transformation and then calculates thresholds in selected time windows, eventually achieving frequency-divided noise attenuation for better denoising. Simulations based on theoretical and real-world data verify the adaptability and low dependence of the method on the size of the time window. The method suppresses noise without energy loss in the signals.
