next up previous [pdf]

Next: Acknowledgments Up: Liu etc.: - - Previous: Field data test

Conclusions

We have introduced a new approach to adaptive prediction filter (APF) for seismic random noise attenuation in $ t$ -$ x$ -$ y$ domain. Our approach uses regularized nonstationary autoregression (RNA) to handle time-space variation of nonstationary seismic data. These properties are useful for application such as random noise attenuation. The predicted signal provides a noise-free estimation of local plane events. Compared with the $ f$ -$ x$ -$ y$ NRNA method, $ t$ -$ x$ -$ y$ APF can capture more detailed signal and avoid most artifacts, which occur more in frequency domain methods. However, the $ f$ -$ x$ -$ y$ NRNA method uses fewer prediction coefficients (no time prediction) to save storage space and can be applied in parallel to different frequency slices. Therefore, $ f$ -$ x$ -$ y$ NRNA is appropriate for mild complex structure and fast computation while $ t$ -$ x$ -$ y$ APF is more appropriate for very complex structures. Experiments with synthetic examples and field data tests show that the proposed filters are able to depict variations in the nonstationary signal and provide a accurate estimation of complex wavefields even in the presence of strongly curved and conflicting events.


next up previous [pdf]

Next: Acknowledgments Up: Liu etc.: - - Previous: Field data test

2014-12-07