Seismic Coherent Noise Removal Based On The Curvelet Transform

Seismic exploration can be divided into three parts: seismic data acquisition,seismic data processing and seismic data interpretation.Among them,seismic data acquisition refers to the process of artificially stimulating seismic waves by blasting and then using a geophone to receive.The actual collection process is limited by natural conditions,or the impact of the acquisition device,etc,the data collected are often polluted by various noise,which seriously affects the subsequent interpretation.Therefore,denoising is imminent.According to the differences of noise propagation law,noise is divided into two categories: coherent noise(non random noise)and random noise.Among them,coherent noise can be divided into linear noise,multiple wave and ground-roll,etc.The research of this paper focuses on the removal of ground-roll and linear noise of seismic coherent noise.The characteristics of linear noise are that the apparent velocity is stable,and the apparent velocity is different from the effective signal.In seismic records,it can be clearly seen that their phase axis angles are very regular;the frequency is close to the effective reflected wave.Curvelet has multi-scale,multi-directional and tight frames,and is approximately linear in time domain.It can optimally represent seismic data sparsely.The seismic wave groups with different frequencies,directions and spatial positions can be well separated by the Curvelet transform.According to the velocity and space position differences between linear noise and effective signal,the seismic data are transformed by Curvelet.Analyzing different features of Curvelet coefficients,selecting the appropriate threshold to Curvelet coefficients,and then reconstructing back,which has a good effect on suppressing linear noise;Comparing with traditional filter,which shows its advantage in suppressing linear noise.Because of the diversity of morphological components and the development of sparse representation theory,morphological component analysis(MCA)theory is proposed.This method has a good effect on the separation of complex signals,so it is widely used in the separation of seismic effective signals and noise.In this paper,the method is applied to ground-roll suppression,and good results are obtained.In the course of the propagation of seismic waves,some of them propagate along the earth's surface,and they interact with each other.This wave is called ground-roll.Ground-rolls are usually "broom like",with strong energy,low frequency,low velocity,and almost vertical direction.In view of the above characteristics of ground-rolls,the Curvelet transform is usually chosen as the sparse representationdictionary of them.Considering the strong local feature of volume waves,local discrete cosine transform(DCT)is chosen to represent them sparsely.Then,the model of ground-roll separation is established by using morphological component analysis(MCA)method.The block relaxation method is used to solve the problem,and the ground-roll is separated from the effective signal so as to obtain the effect of suppressing ground-roll.The method is applied to synthetic data and real data,and then compared with the traditional filter,which shows its advantage in suppressing ground-roll.