Abstract: Traditional Time-frequency Peak Filtering (TFPF) usually causes errors at peaks and valleys of valid signal. This study proposed a parallel Radial Trace Transform (RTT)-TFPF with x-interpolation method to reduce these errors. The proposed method uses TFPF to eliminate noise after mapping seismic signals from coordinates of offset and traveltime to those of apparent velocity and traveltime, namely filtering the signals along the direction of the radial trace in the time-space domain. The effect of the different slope of radial trace on the unbiased estimation of TFPF was discussed. Simulation experiments of synthetic seismogram were carried out to compare the traditional TFPF and the RTT-TFPF with different slopes. The results show that when the selected radial traces and events are parallel or the angles between them are small, the proposed method can make the signals better meet the condition of unbiased estimation in one window length, reduce the errors of the peaks and valleys of filtered signal, and improve the property of amplitude-preservation, the SNR of the proposed method can increase about 5 dB.

Key words: information processing, radial-trace transform, time-frequency peak filtering, random noise, signal enhance