Wiener systems for reconstruction of missing seismic traces

This paper presents a new method for the reconstruction of missing data in seismic signals. The method is based on Wiener systems considering non-Gaussian statistics in the probability density function of the seismic data. Wiener structures are proposed combining different techniques for the linear and non-linear stages. The linearity in the data is recovered using kriging and cross correlation, and the data nonlinearity is reconstructed using direct sample estimation and a third order polynomial approximation. The results by linear and Wiener structures are compared with the results of Multi-Layer Perceptron and Radial Basis Function networks. Several examples with real data demonstrate the efficiency of the method for seismic trace reconstruction. The accuracy of the recovered data is evaluated by the error of the estimates and statistics of the data density for the recovered data.