Seismic Random Noise Attenuation by Time-frequency Peak Filtering Based on Curvelet Transform

Abstract

Time-Frequency Peak Filtering (TFPF) is an effective method to eliminate random-noise and has been applied to seismic noise attenuation in recent years. In the conventional TFPF, we use pseudo Wigner-Ville distribution (PWVD) for instantaneous frequency (IF) estimation. But when the signal-to-noise ratio (SNR) is low, the random noise will have an interference on IF estimation, which results in a bias in TFPF. To solve the problem, we apply curvelet transform to eliminate the noise interference around the IF. The curvelet transform is a new multiscale transform with the optimal directional character that provides an optimal detection to curve signals from strong noise interference. Hence, PWVD processed by curvelet transform realize the removal of interference without the energy loss of IF curve. Then we use the processed time-frequency distribution in TFPF to reduce the bias caused by noise interfer. Experiments on real seismic data demonstrate that the TFPF based on PWVD processed by curvelet transform can suppress random noise more effectively and preserve events better than conventional TFPF.