The F-X prediction filter is a technique to suppress random noise from seismic data in the frequency-space domain. For a given frequency, the coherent signal is predictable and random noise is unpredictable in the F-X domain. We then use the Complex Wiener Filter theory to obtain a least-square approximation of the desired signal.
The method has been tested on a noisy stacked section. The result was acceptable without any noticeable artifact (lateral mixing). The method also applied on a synthetic shot record to eliminate air waves (coherent noise). This required application of an appropriate linear moveout on data before using F-X prediction filter. The necessary codes were written by authors in MATLAB environment