Research On Seismic Data Denoising Method Based On Sparse Representation

With the deepening of geological exploration,the exploration target area is increasingly complex,and the collected seismic data is often interfered by random noise.At the same time,random noise is inevitably introduced during seismic data processing.The random noise and coherent noise contained in the collected seismic data will greatly reduce the signal-to-noise ratio of the seismic data,and will also affect the subsequent analysis of seismic data and the judgment of the geological environment of the target area.Therefore,the noise reduction processing of seismic data has become an important part of seismic data analysis and processing.This paper focuses on the random noise suppression method in seismic data.The main work is as follows:(1)We study the noise reduction method of random noise in seismic data.The noise in the seismic data contains non-stationary features,which cause great trouble for the subsequent processing of seismic data.When the traditional wavelet,Curvelet threshold and other noise reduction methods are used to denoise the seismic data,the effect is not satisfactory.To solve this problem,a seismic data denoising method based on sparse representation is introduced.In this paper,Bidimensional Empirical Mode Decompostion(BEMD)and Shearlet transform are selected as the main methods of noise reduction,and the theory of the two is analyzed to provide theoretical basis for the proposed noise reduction method and experimental design.;(2)We propose a joint noise reduction method based on BEMD and Shearlet transform.Conventional wavelet,Curvelet threshold and other noise reduction methods use only a single transform in noise reduction,and cannot be processed according to the distribution of seismic data noise in each frequency segment,and the noise reduction results contain"artifacts" and noise reduction is not complete and useful information is lost.In order to solve this problem,this paper proposes a joint noise reduction method based on BEMD method and Shearlet threshold method.Firstly,the seismic data is decomposed into a series of IMF(Intrinsic Mode Function,IMF)components with high frequency to low frequency through BEMD;Then,the noise-free high-frequency IMF component is denoised using the Shearlet threshold method.The method takes into account the advantages of BEMD and Shearlet transform,not only effectively removes noise,but also preserves effective information in seismic data and improves the noise reduction quality of seismic data;(3)This paper is to find the threshold point of the noise dominant component and the effective information dominant component in the IMF component after BEMD processing.In this paper,the Fourier spectrogram is used to analyze the spectrogram of each IMF component,and the boundary between the noise dominant component and the dominant component of the effective information is obtained.This paper uses the correlation coefficient after autocorrelation analysis to verify it;(4)When using the Shearlet threshold method to denoise the seismic data,the selection of the threshold has an important influence on the noise reduction result.However,the traditional threshold cannot take into account the relationship between the coefficient scales after Shearlet decomposition,introduce the threshold related to the scale,and reduce the noise.It is related to the scale in the process.By analyzing the experimental data,the peak signal-to-noise ratio and root mean square error of the seismic data after noise reduction are analyzed.It is verified that the above algorithm can better preserve the seismic data features while removing noise.