Multichannel Matching Pursuit for Seismic Trace Decomposition

Matching pursuit can decompose a seismic trace adaptively into a series of wavelets. However, the solution is not unique and is strongly affected by data noise. To improve the stability of the procedure, matching pursuit is implemented in a multichannel fashion that exploits lateral coherence as a constraint to overcome the non-uniqueness of the solution. Each constitutional wavelet extracted from a seismic trace has an optimal correlation coefficient to neighboring traces. The scheme stabilizes performance on wavelet decomposition with great improvement over spatial continuity along a seismic profile. It is demonstrated using two examples. One shows that the multichannel scheme is able to remove a strong coal-seam reflection from seismic profile, before target reservoirs with weak reflections on the top can be characterized. Another is to generate a reliable, spatially continuous time-frequency spectrum, on which low-frequency shadows can be used for gas reservoir detection.