Full text loading...
-
Stagewise Conjugate Gradient Pursuit for Seismic Trace Interpolation
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 77th EAGE Conference and Exhibition 2015, Jun 2015, Volume 2015, p.1 - 5
Abstract
During last decade Fourier methods based on Compressive Sensing sparsity promoting tools have been ascertained as most effective approaches to the interpolation of irregular sampled data problem. This is due to their capability in handling the non-orthogonality of the Fourier dictionary when data sampling is irregular. The assumption that seismic data have a sparse representation in the Fourier domain make possible the usage of greedy algorithms in order to estimate the spectrum while contrasting the energy leakage effect. The well-known Matching Pursuit and Orthogonal Matching Pursuit have been applied with this aim, by the way both of them turned out to be heavily time-consuming approaches. Thus, next challenge is to come out with a method which is not only effective, but also efficient. In this sense the usage of the recently developed Stagewise Conjugate Gradient Pursuit is proposed. It differs from above mentioned approaches in two main innovations: in selecting more than a single component per each iteration and in updating components by conjugate directions at each iteration. Experiments on a real 3D marine dataset show how Stagewise Conjugate Gradient Pursuit converges much faster than Matching Pursuit to the desired solution while not affecting accuracy of recovered data.