Research On Improved Seismic Noise Suppression Algorithm Based On Block Matching

As a means of exploration for traditional resources,seismic exploration has been developed for a long time.In signal acquisition,the seismic wave received by the geophone contains all the fracturing information.However,in the process of seismic data analysis and interpretation,only effective events can provide valuable stratigraphical information.Random noise in seismic data increases the difficulty to identify effective events.Therefore,it is necessary to study a random noise suppression method which can effectively improve the signal-to-noise ratio(SNR)and resolution of seismic data.The main research in this paper is to reduce the random noise in seismic signal.As the hotspot,there are a lot of effective solutions in the field of seismic noise suppression.Among these solutions,the TFPF is an excellent algorithm both in the signal to noise ratio improvement and computational time complexity.Firstly,this paper introduces the fundamental principles of the TFPF algorithm and the PT-TFPF algorithm.It summarizes the condition of the TFPF algorithm for unbiased estimation.Secondly,the advantages and disadvantages of the two algorithms are verified by experiments on synthetic seismic signal and filed seismic record.These analyses and experiments are well prepared for subsequent improvements.By analyzing the condition of the TFPF algorithm for unbiased estimation,this paper proposes an improved method of TFPF based on block matching to meet the demand of signal amplitude preserving.Firstly,the original signal is decomposed into a series of same-sized sub blocks.Due to the cross correlation,the waveform of sub blocks on adjacent channel is similar.By calculating the Euclidean distance,the similarity degree between the blocks is measured.The similar blocks selected by the threshold are rearranged into a matrix by columns.Because columns are similar,the linearity of rows in new matrix is improved significantly.Therefore,unbiased estimate is implemented when the filtering is carried out along the row of the new matrix.Finally,put the estimation of each sub block back to the original position.The seismic data after noise reduction is calculated by the weighted aggregation.Compare to the PT-TFPF,BM-TFPF not only achieves the adaptive linearity improvement,but also reduces the signal attenuation caused by fixed window function.Aiming at the problem of noise suppression in high frequency band,considering the frequency characteristics and the directional features(the effective signal is anisotropic and the random noise is isotropic),this paper proposes an improved method based on block matching in dual-tree complex wavelet domain.Firstly,according to the direction information of the seismic events,this algorithm decomposes seismic signal by multidirectional 2D DTCWT.For the low frequency scales,the signal content in each direction component is judged by the signal energy.And different threshold values are used to suppress the noise in these components.This way eliminates a large number of random noise without losing the effective signal.Secondly,to suppress random noise in the high frequency scales,the block matching algorithm is used to construct the low rank matrix according to the cross correlation of the effective events.Then,the low rank approximation is realized by using the singular value decomposition to achieve the filtered high frequency component.Finally,the various noise suppressed components are reconstructed by dual tree complex wavelet.Based on the study of TFPF and DTCWT,this paper constructs two improved noise suppression schemes,and applies them to synthetic seismic signal and filed seismic record to verify the feasibility and effectiveness.