Deconvolution based on unconstrained forward-backward prediction error minimization
Zong Tao1, Meng Hongying1, Liu Guizkong1, Wu Skuanhu2
Research Institute of Information Engineering, College of Electronic and Information Engineering, Xi'anJaiaotong University, Xi'an City, Shanxi Province 710049
Abstract:Under unreasonable presumption that wavelet doesn't vary with time,processing whole seismic trace by pulse deconvolution results in the seismic section that has high-frequency contents in shallow part but low-frequency contents in deep part.It has been shown that windowing also brings poor effect of the deconvolution when the time window is short.Burg deconvolution,using running window,is suit- able to time-varying property of wavelet;but restricted by Levinson relation,it still fails to cast off the characteristics of Toeplitz matrix and to remove fully the defect of pulse deconvolution.To cope with the problem,we advance the deconvolution based on unconstrained forward-backward prediction error minimization.The method needs neither the solution of Toeplitz matrix nor the presumption that wavelet does not vary with time,and it can be fast achieved by using MARPLE recursion algorithm.Theoretical and real results say that the deconvolution method is firm and effective on both small window and big window.