ITERATIVE ADAPTIVE APPROACH FOR SEISMIC DATA RESTORATION

摘要

Reconstruction of missing traces of seismic data from finite samples is a problem in seismic data processing. In this paper, an iterative adaptive approach is proposed to restore seismic data with randomly missing traces, specifically, which is suitable to recover a large number of missing traces. The proposed method is based upon the weighted least square theory. Unlike previous low-rank methods that use the low-rank property of the Hankel matrix on each frequency slice, we exploit the harmonic structure of frequency slice, and develop an iterative adaptive manner for seismic temporal frequency slices to obtain an accurate spectral estimation. The missing data is filled using a linear minimum mean-squared error estimator. Numerical experiments show that our method provides much better performance for reconstruction compared to that of the classical low-rank methods such as iterative soft thresholding, low-rank matrix fitting and orthogonal rank-one matrix pursuit.