2-D TFPF Based On Contourlet Transform For Seismic Random Noise Attenuation

Seismic exploration technology is one of major techniques of geophysics exploration. Seismic data is not only the basis of mineral resources exploration, but also widely used in the research of earth crust. According to the flexibility discrepancy of rocks, in the seismic exploration, seismic waves are generated artificially and spread in geologic layers, in accordance with which we can estimate geologic structure and serve for mineral resources search. Because of complicated environments in exploration, seismic signals are corrupted by a variety of noises, especially the random noise which disturbs the estimation of geologic structure. Therefore, seismic noise attenuation plays an important role in improving the accuracy of seismic exploration.The time-frequency peak filtering(TFPF) algorithm is a new and useful for attenuating seismic random noise. For an ideal result without filtering error, the high linearity of signal is required in the TFPF. In seismic signal processing, conventional TFPF use a single fixed window length to improve part of signal linearity but a single fixed WL is inappropriate for all frequency components at the same time, leading to serious loss of some effective components. Secondly, it filters the seismic record along the channel direction and thus ignores spatial characteristics of seismic reflection events, resulting in discontinuous seismic events. In order to solve these drawbacks in conventional TFPF, the 2-D TFPF based on Contourlet transform(CT-TFPF) is proposed in this paper.In the filtering of our method, utilizing the cross-correlation characteristics of seismic signals, firstly we regard seismic events as the contour of 2-D signals to find seismic events by using the CT. After that, with the found information of the event, we can get an optimal filtering trace approximately aligning with the event. Through resampling the original record along the optimal trace, the amplitudes of sampled data are similar and the linearity of the input signal of the TFPF has been improved greatly. Finally desired signals can be recovered with less bias by filtering and random noise can be suppressed effectively. We apply the proposed method in the processing of synthetic seismic data and field seismic record to test its feasibility and effectiveness. In the results, it has been shown that our method can suppress random noise to a large extent and preserve the desired signals continuously and completely.