Recursive removal and estimation of minimum‐phase wavelet

Abstract

The Gram‐Schmidt orthogonalization procedure is simplified under the assumption of stationarity and implemented to perform recursive predictive deconvolution. This process is called the orthonormal lattice filter. The results of deconvolution by this method are compared to the results of deconvolution using the maximum entropy and Levinson algorithms. The orthonormal lattice and maximum entropy algorithms are direct methods and estimate the filter from the data, while the Levinson algorithm is an indirect method and estimates the filter from the autocorrelation function. Results from synthetic and real data show that the direct methods have fewer windowing problems, higher resolving power, and are more suited for use in a time‐varying manner than the indirect method. Results from real data show that optimally weighted space averaging and zero‐phase band‐pass filtering after time‐varying direct deconvolution produces highly resolved and spatially coherent seismic sections.