Abstract:The mixed-phase seismic wavelet estimation approach based on high-order statistics developed in resent years often assumed that the reflectivity sequence of strata is non-Gaussian and statistically independent random process rather than Gaussian white noise distribution, which is more consistent with practical situation. The approach presented in the paper first used trispectrum phase information of high-order statistics on the seismic traces to reconstruct phase spectrum of seismic wavelet, and yielded the amplitude spectrum of wavelet by autocorrelation of seismic traces, from which the seismic wavelet with mixed phase could be detected, realizing the deconvolution of seismic traces by improved Cauchy-constrained maximum posteriori deconvolution. It is shown by simulation tests and practical seismic data processing that maximum posteriori deconvolution can improve the resolution of seismic data in a precondition without reducing S/N ratio of seismic data.