Abstract:Having changed a two-variate seismic signal into a four-variate function relating to time, frequency, space and wavenumber with the use of continuous 2D wavelet transform, we can suppress many noises (such as surface waves and regular noises) to enhance the signal/noise ratio of seismic data by making full use of the differences between signal and noise in four aspects:time, frequency, space and wavenumber. In the w-x domain for wavelet transform. signals can be predicted in space direction while noises can not, so that random noises can be suppressed through prediction. The method is superior to f-x domain prediction. Real data processing proves the method feasible.