Research On Seismic Noise Suppression Method Based On Shearlet Transform

In seismic exploration,the seismic data usually acquired in the field are mixed with a variety of noises.On the one hand,external natural conditions and artificial sources constrain the signal acquisition environment,and on the other hand,current disturbances are inevitably introduced during the recording process of sensors.The presence of these noises seriously reduces the signal-to-noise ratio and resolution of the seismic profiles,and brings difficulties to the subsequent seismic data processing and interpretation work.Therefore,it is especially important to suppress the noise and improve the quality of seismic data.In recent years,with the improvement of computer performance,multiscale geometric analysis methods have been greatly developed,which can extract the information of images on deep structure and improve the accuracy of image feature description.From the initial wavelet analysis,which expands the limitations in the direction of one-dimensional representation,to the present day,various multiscale geometric analysis methods represent high-dimensional data sparsely and effectively,and multiscale geometric analysis is still under development,and new sparse representation algorithms are expected to be useful in the field of seismic signal processing.The suppression of random noise involves two main aspects: the effective separation of signal and noise and the edge preservation.In this paper,we first improve the Shearlet transform combined with the bivariate shrinkage method for both high and low frequencies,and analyze the similarities and differences in noise suppression for high and low signal-to-noise ratios,and conclude that the Shearlet shrinkage denoising method for low signal-to-noise ratios brings about the "cracking" phenomenon,while combining the Shearlet transform The combination of Shearlet transform and non-local mean method can effectively overcome this problem.In addition,we propose an adaptive Shearlet transform threshold denoising method with scale for batch processing of seismic data,and compare the two-dimensional and three-dimensional Shearlet transforms.The comparison tests show that the 3D Shearlet transform can suppress the strong amplitude interference while better protecting the homogeneous axial information under complex structures.