Abstract:Compared with the routine deconvolution,the spectral modeling deconvolution has better adaptivity for non-white noise reflection coefficient,with the method the better results can be obtained in the processing in which the residual wavelet was compressed so the resolution was raised.However the method has following limitations:(1) The polynomial used in fitting limits the shape of the wavelet amplitude spectrum;(2) It does not consider the color characteristics of the reflection coefficient,the expected blue characteristic of the reflection coefficient amplitude spectrum was not seen in the deconvolution results.In order to compensate the defects,a new way to realize spectral modeling deconvolution was proposed in this paper:(1) Do not use polynomial fitting,regarding the amplitude spectrum of the seismic records as a real sequence,a filter was used to smooth the sequence so the wavelet amplitude spectrum could be estimated;(2) Considering the color characteristic of the reflection coefficient sequence,before modeling the wavelet amplitude spectrum,the color influence of the reflection coefficient sequence was compressed.The theoretical experiment and processing results for the field datasets show that the method is simple and easy to realize,with the method the wavelet amplitude spectrum can be accurately estimated,better deconvolution results can be obtained.