Application Of Matching Pursuit Algorithm In Seismic Data Processing

Matching pursuit algorithm is a powerful tool in processing unstable signals.The signal can be decomposed into a series of linear combinations of atoms by it.The purpose of this paper is to apply this algorithm into non-stretching NMO,time-frequency analysis and pre-stack elastic impedance inversion.For non-stretching NMO,we apply matching pursuit to correct the time shift of a single atom.By doing so,we can avoid correcting events point by point and prevent seismic data from stretching.In this paper,we add the lateral constraint to this algorithm to keep continuity of corrected events and speed up the calculation by matching pursuit of complex field.For time-frequency analysis,compared with other traditional methods which are affected by window function,matching pursuit with Wigner distribution for signal time-frequency spectrum has a higher resolution.But wavelet side lobes may increase the extra energy to the spectrum.In order to eliminate effects of wavelet side lobes,we add the Gauss attenuation function to reduce the side lobes'energy and improve the resolution of signal time-frequency spectrum.For the inversion of elastic impedance,inversion results which are calculated by the least square method are unstable and have a low resolution.In this paper,we add the constraint of impedance regularization to the objective function to enhance the stability of inversion and utilize L₀ norm to constrain the sparsity of reflection coefficients.In this way,we can obtain blocky elastic parameters with higher resolution of layers.