Adaptive method for swell noise attenuation in the time-frequency domain
Julio Valença Tavares, Luiz Geraldo Loures · 2009
Coherent noises are those that appear through a systematic way and swell noise is a kind of coherent seismic noise. This happens during the marine seismic acquisition and is generated basically by the motion of the ship, cable depth controllers, or sea surface waves. For the swell noise analysis and attenuation we used a one dimensional wavelet transform, in the sense that it operates on single traces, to apply a time-frequency domain filter. The shifting (time) and level (frequency) dependent soft-threshold is estimated using a method based on Bayes’ Rules. Using this proposed methodology for synthetic data analysis the results were satisfactory. When real seismic data is used the results show good agreement due to the algorithm adaptive structure. The noise is removed considerably and the signal doesn't loose the interesting information, validating the proposed model. As the algorithm makes the threshold estimation in function of the scale and for each trace time interval, nine threshold values are calculated for each trace. Therefore, it is possible to reduce the noise for different signal-to-noise ratios (SNR) along each trace. The noise free traces don't change, because in those cases the threshold tends to be zero making possible the perfect signal reconstruction. However, when the SNR tends to be one or the noise and signal amplitude and frequency are similar their coefficient variances tend to be close causing a threshold estimation enough to mute the signal. In agreement with the increasing demand for offshore exploration, also driven by 4D seismic acquisitions, the swell noise will be a frequent problem, because in order to achieve a good geologic model all the distortions generated by the data acquisition should be attenuated.