Noise Attenuation Based On Inversion And Software Development On The Geoeast Environment

Noise attenuation is an important work throughout the whole process of seismic data processing and an important means to improve the signal-to-noise ratio of seismic signals.In order to meet the requirements of quantitative seismic data interpretation,the noise attenuation method should also protect the effective signal while suppressing the noise and maintain the relatively complete dynamic characteristics of seismic reflection.Noise reduction methods based on predictive filtering(such as f-x deconvolution)are one of the most commonly used random noise suppression methods in the industry today.However,the f-x deconvolution has two different assumptions for the noise model.One is that the seismic record is summed by the effective signal and random noise,which is the so-called additive noise model,and the other is that the random noise is the convolution of the prediction error filter and the seismic data,which is the so-called source noise model.Two different noise models reduce the theoretical rigor and adaptability of the method,weakening its the denoising and signal-preserving ability.For this reason,this paper proposes a random noise attenuation method based on inversion.It uses the predictive filter operator as a reflection structure constraint to introduce the inversion system,and inverts the seismic signal from the seismic record,avoiding the inconsistency of the noise model.This way enhances the denoising and signal-preserving ability.Due to the complexity of the underground stratigraphic structure,the real seismic data is non-stationary obviously.For this reason,we extended the method to the denoising processing of non-stationary seismic signal,and proposed the non-stationary seismic signal inversion method based on the shaping regularization.The methods were developed on the GeoEast environment and passed third-party tests.The application of model data and real data shows both methods have better noise attenuation and signal protection capabilities than predictive filtering methods.