The Application Of Diffusion Filtering Methods In Seismic Exploration

In seismic data processing, people usually require high signal-to-noise ratio,high-resolution and high-fidelity real seismic data. Therefore the de-noising methodwhich is used to seismic data has been a hot topic in the field of seismic dataprocessing. Over the years, people design a variety of de-noising methods and itsapplication in the seismic data processing flows to suppress the different types ofnoise, and these methods achieved good results. Conventional de-noising technologyof seismic data has been increasingly unable to meet the needs of the development ofseismic exploration technology at present, the limitations of the traditional de-noisingtechnology resulting in the de-noising technology of seismic data continues todevelop.The diffusion filtering method is a kind of the image processing method which isbased on partial differential equations (PDE).1990s, the PDE-based image processingmethod is a rapid development of cross-disciplinary branch. Compared with thetraditional image processing method, diffusion filtering method has obviousadvantages in terms of image noise reduction, image segmentation, imagereconstruction, edge detection and image restoration, it is more suitable for lineartexture visible image, such as images of medical, images of seismic exploration,graphics and other images. The PDE-based image processing method originated fromthe heat propagation and diffusion in Physics, this method can use the image toportray the distribution characteristics of the heat source at different times. Thismethod regard the original image as the initial input value, it solve the partialdifferential equations which satisfy the image after a constant diffusion time. The process of diffusion can be controlled under some regular diffusion according to thelocal image characteristics.In order to get a high signal-to-noise ratio, high-resolution and high-fidelity realseismic data, this paper mainly studies the four kinds of diffusion filtering methods.This paper discuss about theory of methods, foundation of mathematics and algorithmof programming of the four kinds of diffusion filtering methods in detail whichinclude the nonlinear isotropic diffusion filtering method, the nonlinear anisotropicdiffusion filtering method, the complex domain diffusion filtering method and thewavelet modulus maximum time-frequency domain diffusion filtering method. In thispaper, propose to make some improvements on some algorithm at the sometimes. Thenonlinear anisotropic diffusion filtering method on the basis of Weickert’s nonlinearanisotropic diffusion filtering method, uses Jacobi method to build structure tensor Dand solve its eigenvectors and eigenvalues. A linear confidence measure C is built tosolve the coefficients. And then calculate the diffusion filter coefficients. Finallysubstitutes the diffusion filter coefficients into nonlinear anisotropic diffusionequation directly. In the study of the wavelet modulus maximum time-frequencydomain diffusion filtering method, the choice of the wavelet function is different fromprevious Haar wavelet function, this method use Molert wavelet function. This paperuses the Aki reflection coefficient approximate formula fitting out the four categoriesAVA curves, and performances angle1o-45oseismic prestack AVA angle gathers.This paper regard prestack AVA angle gathers as the algorithm experimental modeldata to analyze the effect of this paper of the types of diffusion filtering method tosuppress the noise. This paper also discuss the diffusion filtering method is integratedin the Geoscope2.7software platform which is developed by Chengdu RockstarPetroleum Technology Co., Ltd., and then the diffusion filtering method serve theactual production projects. Finally, the diffusion filtering method is applied to the realseismic data, according to the noise suppression effect of the prestack CRP gather, thenoise suppression of the2-D post-stack seismic section with low signal-to-noise ratioand the noise suppression effect of the2-D post-stack seismic section containingcomplex geological structure information to select the most appropriate diffusionfiltering method integrated in the Geoscope2.7software platform In this paper, thenonlinear isotropic diffusion filtering method and the complex domain diffusionfiltering method are not suitable for suppressing random noise, apparently randomnoise is not suppressed. The nonlinear isotropic diffusion filtering methods suppressthe horizontal and vertical directions with the obvious continuity noise effect. The nonlinear anisotropic diffusion filtering method is equivalent to a directionalinclination filtering method, it can keep the continuous seismic event and the shape ofseismic event, it can remove linear or inclined linear noise. Wavelet modulusmaximum time-frequency domain diffusion filtering method also has the advantage inkeeping the seismic event, it`s effect of the removal of random noise significantlybetter than both effect of the nonlinear isotropic diffusion filtering method andcomplex domain diffusion filtering method.