Seismic Blind Deconvolution Algorithm Based On Bayesian Framework

Blind seismic deconvolution procedure need to be applied to recover the unknown seismic wavelet and the reflectivity sequence.The purpose of blind seismic deconvolution is to improve the resolution ratio of seismic reflectivity sequence,broad the spectrum effectively and keep high frequency component.As a result,accurate stratigraphic structural characteristics can be obtained.Seismic blind deconvolution algorithm is widely used in seismic signal analysis,geological investigation,resources exploration,marine seismic exploration and other fields.Seismic blind deconvolution problem based on Bayesian framework is studied in this paper.The prior models for the seismic wavelet,reflectivity and the seismic record are established.Parameter estimation methods are used in the algorithms.The main work is as follows:(1)Seismic blind deconvolution based on variational Bayesian and seismic blind deconvolution based on sparse representation variational Bayesian are proposed.The prior models for the seismic wavelet,reflectivity and the seismic record are established.Iterative formulas in the algorithms are deduced.The computer simulation shows that the new algorithm obtains high precision reflectivity sequence and reduces the mean square error.(2)Seismic blind deconvolution algorithm based on Gibbs sampling Bayesian multichannel is proposed.Markov Bernoulli Gaussian model is used as the model of multichannel reflectivity sequence.Maximum posterior distribution of the reflectivity sequence is approximated by the partially collapsed Gibbs sampling.Reflectivity sequence is abtained.Furthermore,Markov Bernoulli Gaussian model is improved and then seismic blind deconvolution algorithm based on partially marginalized Gibbs sampling Bayesian multichannel is proposed in this section.Experimental result shows that the new algorithm broadens the effective spectrum of seismic reflectivity sequence and reduces the loss function.(3)Blind deconvolution algorithms based on Bayesian compressed sensing using linear wavelet and Curvelet transform are proposed respectively.The prior models based on Bayesian compressed sensing for the seismic wavelet,reflectivity and hyperparameters are established.The expectation maximization strategy is used to achieve parameter estimation.The computer simulation shows that the new algorithm reduces the normalized mean square error and improves the estimation precision of seismic reflectivity sequence.