The Research Of Seismic Random Noise Suppression Method Based On The WNNM

Seismic exploration technology is an effective method to obtain information of mineral resources distribution.The seismic waves excited by certain techniques often carry the geological information we need.These information can help us to understand the geological structure and locate the mineral resources.With the continuous reduction of mineral resources,seismic exploration needs to be carried out in a more complex environment.The difficulty of seismic exploration is greatly increased.Due to the influence of environment and instrument,the seismic wave is accompanied by a large amount of random noise,which makes it difficult to analyze and explain the seismic data.Therefore,it is the primary task of seismic signal processing to suppress seismic random noise,so as to improve signal to noise ratio.The Weighted Nuclear Norm Minimization(WNNM)algorithm is the improvement of Nuclear Norm Minimization(NNM)algorithm.The algorithm is first used for image denoising.It uses the non local similarity of the image to build low rank matrix.Then through the low rank approximation can achieve the reduction of noise.Compared with NNM,WNNM shrinks the singular values according to their amplitude.It considers that the larger singular values contain more effective signals and should be shrink less to retain the effective signal.In this paper,the method of low rank matrix approximation based on WNNM is introduced into the seismic random noise suppression,aim at to suppress the random noise and better protect the effective signal.Seismic events are similarity in time domain and space domain.Considering this characteristic,in this paper,the non local filtering based on the WNNM is used to process the seismic signal.Firstly,the seismic signal is decomposed into a series of seismic texture blocks with the same size and overlapping.For each block,the block matching is used to obtain its similar seismic blocks in the neighborhood.Then these blocks are stretched and arranged according to the similarity to construct matrix.Due to the similarity between the matrix columns,the matrix is approximately low rank and it can be filtered by the WNNM low rank matrix approximation.Finally,each column in the matrix is converted into a two-dimensional block and then put back into the original position to form the denoised seismic signal.In order to achieve better filtering effect,the seismic signal is processed by iterative filtering.We use this method to process the synthetic seismic data and the real seismic records.It can suppress the random noise while retaining the effective signal.The WNNM method use the same noise variance to process different matrices,which ignored the noise intensity of seismic signals in different regions and cannot deal with the matrices differently.To solve this problem,in this paper,we propose an adaptive WNNM algorithm based on Complementary Ensemble Empirical Mode Decomposition(CEEMD).Firstly,the seismic signal are decomposed by CEEMD,and the first component is the noise component.Then the local noise variance of seismic signal is calculated from this component,and it is used to set the weights.This method improves the disadvantage of WNNM using the global noise variance to set the weights,and realizes the differential processing of different similarity matrix.At the same time,it realizes the Adaptive filtering for different regions of seismic signals.In order to verify the effectiveness of the proposed method,the Gauss white noise with different signal to noise ratio(SNR)is added to different regions of the simulated seismic signal.The experimental results of this method are compared with WNNM.The comparison result shows that the improved algorithm is more effective in suppressing noise when the noise intensity is different in different regions.And the signal-to-noise ratio is improved by about 3d B.In addition,the processing result of the actual seismic records further verifies that the proposed method improves the applicability of WNNM.