Seismic Signal Denoising Based On Sparse Prior

The signal denoising of seismic exploration is the key step to improve the signal-to-noise ratio of seismic signal,and it is also a very challenging research problem in the field of seismic signal processing.In recent years,due to the sparsity of seismic signals in the transform domain,seismic signal processing based on sparse prior has gradually become one of the research hotspots in the field of the exploration of the seismic signals.In this paper,the research of seismic signal denoising is carried out,which is based on the framework of sparse prior of seismic signal.The main contents of this paper are as follows:(1)Seismic signal denoising based on region segmentation gradient histogramThe gradient prior of seismic signal belongs to the local sparse prior.Based on this,a denoising algorithm based on SGHP is proposed.The algorithm firstly divides the noisy seismic signal into several regions,then estimates the reference gradient histogram of each region,finally it uses the algorithm of Gradient Histogram Preservation(GHP)to denoise each region,so that the gradient distribution of the denoised signal is as close to the gradient distribution of the original signal as possible,so as to protect the details of the seismic signal.Denoise the synthetic seismic signal and the post-stack land seismic signal respectively,and compare SGHP with these algorithms: non-local mean filtering(NLM),block matching 3D cooperative filtering(BM3D),clustering sparse representation.The indexes such as peak signal-to-noise ratio and structural similarity are used for evaluation.The results show that SGHP has the best effect in the above denoising algorithm.(2)Seismic signal denoising based on wavelet-high order total variation with overlapping group sparsitySeismic signal denoising based on Total Variation(TV)can denoise the seismic signal in the sparse transform domain.Therefore,a denoising algorithm based on wavelet-high order total variation with overlapping group sparsity is proposed.Firstly,the noisy seismic signal is decomposed into the two parts of the high-frequency part and the low-frequency part by wavelet transform,then the high-frequency part is processed by the algorithm of the hard threshold,the low-frequency part is processed by the algorithm of the high-order total variation with overlapping sparsity,and finally the denoised seismic signal is output by the inverse wavelet transform.Through the denoising experiments of post-stack and pre-stack seismic signals,it is verified that the algorithm in this paper has better denoising effect and can retain the details of seismic signals better than other algorithms such as HOGS-TV,HOTV and OGSTV.(3)Denoising of low-rank approximate seismic signal based on clusteringIn this paper,a low-rank approximation denoising algorithm based on clustering is proposed.Firstly,the seismic signals are clustered throug the algorithm of Gaussian Mixture Model(GMM),and then low-rank approximation is applied to each cluster of signal block and all signal blocks are combined into a complete denoised seismic signal.The low-rank form of seismic signal is also a performance of the sparsity of seismic signal.In essence,the algorithm combines the sparse prior and self similar prior information of seismic signal for denoising.The blocks of the noisy signals are processed by the algorithm of the gaussian mixture clustering,and then the blocks of the same class are processed by the algorithm of the low-rank approximation.This algorithm is compared with other denoising algorithms,such as EPLL,WNNM,PGPD.The comparison of evaluation indicators such as peak signal-to-noise ratio and structural similarity can verify that this paper's algorithm is superior to the above-mentioned denoising algorithms.