Abstract:The paper uses the S transform for time-frequency filtering. Firstly, we introduced in detail the theoretical basis and concrete implement process of S transform, then, compares the achieved results of S transform with that of other filtering approaches, and finally uses S transform to carry out time-frequency filtering of practical data, demonstrating that the time-frequency filtering to denoise by S transform algorithm overcame the shortage that the filtering factors in traditional filtering and denoise approaches can not change with time and frequency variation. For the seismic data that the frequencies of significant events on different horizons varies with times, the shape of time-frequency windows of S transform varies with frequency, that is the width of windows varies with frequency in opposite direction: wide windows in low frequency band, yielding higher frequency resolutions; narrow windows in high frequency band, yielding higher time resolution. It is shown by the application of theoretical records and practical VSP data that time-frequency filtering approach by S transform can not only filter the noises with similar frequency contents and distributing in different time range (such as tube wave), but also diminish the noises that the apparent velocities approach to the effective wave velocities, fully demonstrating the superiority of S transform in the time-frequency filtering aspect.