The Application Of Wavelet Transform And Anisotropic Diffusion Filtering In Seismic Data De-Noising

In seismic data preprocessing, it is an important segment for noise removal andimproving signal-to-noise ratio and resolution, so the research on seismic datade-noising method has been always the key in seismic data processing field. Over theyears, people have worked out many de-noising methods according to the differencebetween effective seismic signal and noise, in which wavelet transform andanisotropic diffusion filtering are two kinds of useful methods.Wavelet transform is a time-frequency analysis method which has multi-scaleand multi-resolution character, which can better describe local characteristics of signalin time and frequency domain. The paper systematically introduces the basic theory ofwavelet transform, multi-resolution analysis and wavelet decomposition andreconstruction and emphasizes on analyzing the application of wavelet threshold inseismic data de-noising. First the article constructed an underground medium layermodel that contained random noise and then adopted low-pass filtering and waveletthreshold to remove its noise respectively. The result shows that the wavelet thresholdmethod can better remove random noise and more improves signal-to-noise ratio andresolution than low pass filtering. After that, it applied wavelet threshold method inpractical pre-stack seismic data to test its effect. We can find the wavelet thresholdmethod has removed the most of the random noise and make the lineups becomeclearer and improve the resolution.Anisotropic diffusion filtering can remove noise quickly and protect the texturecharacteristics when it processes the noise image containing texture characteristics.Since most of the underground media affected by the sedimentary environment, itbasically presents the characteristics of the layer structures. So we can adoptanisotropic diffusion filtering method to remove random noise in the seismic data and protect its edge structure. The paper elaborates the basic principle, equationscalculation and several anisotropic diffusion filtering models and analyzes the role ofdiffusion tensor when filtering processing. Because of the anisotropic diffusionmodel is very sensitive to the default parameters, the paper focuses on modelparameter setting in the anisotropic diffusion and gives a set of parameter settingwhich is fit for anisotropic diffusion filtering of seismic data. At last the paper adoptsrespectively2D and3D seismic data to verify the anisotropic diffusion filteringde-noising effect. The result shows anisotropic diffusion filtering can effectivelyremove random noise as well as protect the edge structure in seismic data and enhanceits continuity. The quality of seismic data has been improved significantly through thecomparison.Based on the research of wavelet transform and anisotropic diffusion filtering,the paper presents a seismic data de-noising method which combines wavelettransform and anisotropic diffusion filtering. The method decomposes the originalseismic data using wavelet, and then uses anisotropic diffusion filtering on lowfrequency and horizontal high frequency wavelet data to enhance the lineupscontinuity and adopts wavelet threshold shrinkage method to process the vertical anddiagonal high frequency wavelet image. This method overcomes the shortcoming ofusing wavelet threshold only removing noise but not enhancing the lineups continuityand using anisotropic diffusion filtering alone can’t de-noise well in much noisecondition. The combined method can remove most of random noise and enhancelineups continuity in seismic data at the same time. Finally, it applies the threemethods to compare their effect with real pre-stack seismic data. The result shows thecombined method has a greater advantage in improving signal-to-noise ratio andresolution and obtains more satisfactory results than using wavelet threshold alone oranisotropic diffusion filtering.