We introduce a fast approach to adaptive PF for random noise
attenuation in the
-
domain. Our approach uses neighboring
similarity of PF to handle time-space variations in nonstationary
seismic data, which is based on elementary algebraic operations and a
streaming method instead of an iterative strategy. Compared with
-
deconvolution and
-
RNA methods,
-
SOPF can
capture a reasonably detailed signal while
avoiding artifacts that occur in the frequency-domain
method. Moreover, the
-
SOPF is superior in terms of its
computational costs. Experiments with synthetic examples and field
data demonstrate that the proposed method is effective at attenuating
the random noise in nonstationary seismic data even in the presence of
curved and conflicting events.