Multi-dimensional Seismic Data Denoising Based On Patch

Noise has a great influence on seismic data acquisition and processing,which leads to the inadequacy of seismic data resolution and signal-to-noise ratio,so seismic data denoising is of great significance.By dividing seismic data into blocks,then grouping seismic data blocks and then denoising,the details of seismic data can be better processed,and the denoising effect is more obvious.Based on seismic data blocks,this paper conducts block-based multi-dimensional seismic data denoising research.The research contents are as follows:(1)Three dimensional seismic data denoising of 4D collaborative filtering block matching(BM4D)combined with wavelet transform researchBM4D is a good three dimensional seismic data denoising algorithm,and this thesis proposes a 3D seismic data denoising algorithm of BM4D combined with wavelet transform to improve the denoising effect.The basic idea is the algorithm preprocesses the seismic data with the wavelet denoising,and integrates the denoising result into the basic estimation part of BM4D.On the other hand,the noisy seismic data is directly input to the final estimation part of BM4D for subsequent wiener filtering and block estimation value aggregation.The synthesized three dimensional seismic data and the actual seismic data are denoised and compared with the wavelet transform and the BM4D denoising results respectively.The experimental results show that the algorithm is feasible.By comparing the output signal-to-noise ratio and root mean square error after denoising,the performance of the algorithm is optimal,followed by the BM4D algorithm,and again the wavelet threshold denoising algorithm.(2)Four-dimensional block matching cooperative filtering 3D seismic data denoising based on principal component analysis noise estimation researchAlthough BM4D has good denoising effect in seismic data denoising,it needs to predict the standard deviation of noise.Regarding the issue above,this paper presents a BM4D three-dimensional seismic signal denoising algorithm combined with PCA noise estimation.Firstly,PCA is used to estimate the noise of seismic data,and then uses the estimation result for BM4D denoising.The experimental results show that the algorithm is feasible,which can not only get a good denoising effect,but also avoid the sensitive limitations of noise level estimation.Compared with other five noise estimation algorithms,the experimental results show that BM4D combined with PCA has advantages in both noise estimation time and accuracy.Therefore,the denoising algorithm proposed in this paper is feasible and superior for denoising 3D seismic data with unknown noise standard deviation.(3)Two-dimensional seismic data denoising based on group sparse residual constraints researchIn this paper,the group sparse residual constraint theory is applied to seismic data denoising research.Inspired by the idea of digital image denoising,this paper proposes a seismic data denoising algorithm based on group sparse residual constraints.In order to reduce the residual,the algorithm maps the seismic data into seismic images.Firstly,the sparse coefficients of the original seismic images are estimated to reduce the residual,then,the coding coefficient closest to the original set of sparse coding coefficients is obtained by processing the noisy seismic image,so that to realize seismic image denoising,finally,it is mapped back to seismic data.The denoising experiments on two-dimensional synthetic seismic data and actual two-dimensional seismic data show that the seismic data denoising algorithm based on group sparse residual constraints is feasible.Compared with wavelet transform(Wavelet),curvelet transform(Curvelet)and three-dimensional block matching filter(BM3D)denoising algorithm,the peak signal-to-noise ratio and mean square error of the proposed algorithm are optimal.Therefore,the seismic data denoising algorithm based on group sparse residual constraints is feasible and superior.