基于互信息率的子波估计与盲反褶积方法研究

西安交通大学硕士学位论文STRACTDeconvolution can improve the vertical resolution of seismic data and characterize thestructure information of stratigraphic,so it becomes a key state during seismic dataprocessing.The blind deconvolution(BD)method based on Mutual Informaion Rate(MIR)criterion is studied in this thesis,Since MIR includes both amplitude information and phaseinformation of random sequence.Firstly,BD both in time domain and frequency domain is investigated.Time domaionmethod has a high computation efficency but a low noise suppression ability.In frequencydomaino method,a constrain term,called noise-correction term,is added in the objectivefunction,to weaken the high frequency interference.And in order to reduce calculation,anew method of initial value selection is proposed,it can decrease the iteration number ofoptimization algorithm about 10 times by contrast with original method.And then,the extraction of band-limited wavelets in noisy record is realized.On the basis ofanalyzing the reasons of noise-sensitive for MIR creterion,a robust technology for seismicwavelts extracting is studied through modifying the noise-sensitive term in the objectivefunction.The shortcome of this method is the enormous amount of computation.Finaly,a novel method for timy-varying wavelets estimation is proposed.First the wavelet'smagnitude spectrum is fitted form the record's magnitude spectrum by using the least squaremethod,and then zero-phase wavelet can be calculated.Then the nonstationary phase isestimation by using the improved Fomel and Baan's work through modifying theregularization constraint term.Finaly the time-varying wavelets can be gotten by making thephase rotation for zeor-phase wavelet with the estimated nonstationary phase.KEY WORDS:MIR;Local Kurtosis;Blind Devonvolution;Wavelet EstimationTYPE OF THESIS:Application FundamentalsSupported by the Major Projects in the National Science Foundation of China(No.40730424),theNational 863 Program of China (No.2006A09A102-11),and Important National Science TechnologySpecific Projects (No.2008ZX05023005-005).