Research On Double-domain Denoising Algorithm For Seismic Data

Seismic exploration is the main technology of global oil and gas exploration.It artificially excites and records seismic waves,and uses computer to process and transform the collected seismic records to obtain structural distribution information of underground media.There is a large amount of random noise in the seismic records measured in the field,which leads to the reduction of the resolution of the seismic records and makes the subsequent data processing unable to achieve fidelity.Therefore,it is necessary to take effective measures to reduce the random noise in the seismic record,restore the original "clean" geological structure,and at the same time preserve as many details as possible such as faults and sharp points of the geological structure.Filtering technology can remove noise,and conventional single domain filtering is usually difficult to effectively meet the de-noising needs of suppressing noise and protecting details.Aiming at reducing noise in seismic records with low signal-to-noise ratio(SNR),this paper studies the dual-domain seismic noise reduction algorithm in both temporal spatial domain and transform domain in order to better protect geological structure information while improving SNR.The main research work of this paper is as follows:1.A seismic denoising algorithm is proposed which is based on improved threshold BM3D(three-dimensional block matching filtering).When BM3 D is applied to seismic records with low signal-to-noise ratio,the low-frequency contour and detail information of seismic records cannot be well protected.In this paper,the threshold function of BM3 D is improved to make it closer to the variation of the wavelet coefficients of effective signal and noise.The experimental results show that the improved threshold BM3 D algorithm is superior to the traditional BM3 D algorithm in both filtering noise and protecting detail information.2.A seismic denoising algorithm is proposed based on anisotropic diffusion correction BM3 D.The main idea is to analyze the BM3 D denoising result specifically.Using wavelet decomposition,the BM3 D denoising result is decomposed into a low frequency and three high frequencies.The denoising effect of each sub-band is verified by experiments,and the frequency band with larger error is found to be corrected by the diffusion in the opposite direction.Experimental results show that the proposed algorithm can not only better protect the details,edges and other information of the seismic image,but also objectively evaluate the signal-to-noise ratio and mean square error value are better than BM3 D algorithm.3.A seismic dual domain denoising algorithm is proposed based on Cycle Spinning and Weighted Nuclear Norm Minimization(WNNM).The dual-domain image denoising(DDID)is easier to implement than BM3 D and the denoising effect is also excellent.This paper introduces the dual-domain image denoising into the field of seismic record denoising.The guided image itself used by the traditional DDID has a large noise residue,which causes the low-frequency component after denoising to be unsmooth,and the Gibbs phenomenon is easily introduced in the denoising process,and there is jitter at the sudden change point.In this paper,the guided image is obtained by Weighted Nuclear Norm Minimization,and the Gibbs problem is solved by the Cycle Spinning.The experimental results show that the two are integrated into the DDID,and the details are in terms of detail protection ability,signal-to-noise ratio and calculation speed.Better than other related algorithms.