Adaptive filtering of random noise in near-surface seismic and ground-penetrating radar data

Adaptive filtering is an effective method to suppress speckle noise in 2D digital image data. Recently, a variety of adaptive filtering algorithms have been developed and employed to remove random noise from geophysical data. In this paper, two filters are designed by adopting adaptive algorithms, the optimum 2D median filter, (a 2D median filter with an optimum window size), and the 2D adaptive Wiener filter (a real time optimal filter renovated from the conventional Wiener filter technology) to investigate the advantages of using adaptive filters in processing ultra-shallow seismic and ground-penetrating radar data. Synthetic common-shot record with added white Gaussian noise was employed to test the effects of 2D window size on both filtering processes. To demonstrate the practical performance of the filter, we processed a set of prestack ultra-shallow seismic data recorded from a shallow fault zone and a stacked section of ground-penetrating radar data as real examples. Examining the performances of the two filters both in time and frequency domains, we notice that the recovery of the original signal depends on the attribute, intensity, and density of the noise. Inspecting the filtered synthetic records in t-x domain, these two filters not only successfully remove the random noise but also suppress the ground roll. With the prestack seismic field data, the median filter renders better resolution than the Wiener filter, but it also suppresses signals that may have geological implications, making the result less desirable. In addition, both the adaptive filters improve the geologically interesting low frequency components of the stacked ground-penetrating radar data, but the high frequency components are blurred.