Iterative deblending of simultaneous-source seismic data using seislet-domain shaping regularization

Conclusions

We have proposed a novel iterative framework, which offers a flexible way to control deblending using sparsity or coherency constraints. When the seismic data are relatively clean and do not contain much coherent noise, $f-k$ domain sparsity-promoting operator or $f-x$ predictive filtering might be adequate to handle the blending noise. However, when the blended data become more complicated, a more robust sparsity or coherency-promoting tool should be utilized. The seislet transform has an inherent ability to compress coherent seismic data, because it is generated based on predictability between neighboring traces using the local slope information. As long as we can get an acceptable slope estimation, the seislet-transformed data appears sparse. Our experiments indicate that it is possible to get accurate results within a small number of iterations when an appropriate shaping operator is taken.