Uma comparação da regularização e interpolação de dados sísmicos 2D com MWNI e Matching Pursuit

Abstract

In seismic data processing, the data are irregularly sampled and often with missing traces, due to logistic restrictions in the geophones and sources distribution. Therefore a prior regularization of this data is necessary, so that we can continue the other processing techniques. These irregular data can be manipulated in the Fourier domain and then interpolated to regular grid by an inverse Fourier transform, since a correct estimate of Fourier coefficients efficiently reconstructs the input data. In this work will be presented the MWNI (Minimum Weighted Norm Interpolation) and Matching Pursuit seismic regularization methods in the simplest cases, considering the interpolation in only one spatial dimension. Then, the numerical results obtained by interpolations with one and two spatial dimensions will be shown. Then, the results of the reconstructions of a seismic section (one shot of the Marmousi data) with both methods in the 2D regularization (with one spatial dimension) will be shown, and finally, the 3D regularization (with two spatial dimensions) of a complete seismic data of the Marmousi and a real land seismic data. We will show the advantages of 3D seismic regularization in relation to 2D regularization, and then a comparison will be made between the methods. In the numerical tests it will be shown that MWNI is faster than Matching Pursuit, but the last one produces slightly better results.