Iterative deblending using the POCS algorithm in the approximate flattened domain

Abstract

Abstract We proposed an improved method to eliminate the interference generated by simultaneous-source acquisition, which can help shorten the acquisition period and improve the quality of seismic data. An iterative mathematical framework is devised, which uses the projection onto convex sets algorithm to estimate the blending noise subtracted from the pseudo-deblended data to separate the blended data in an iterative way. Differently to the conventional method using the coherent-promoting operator only based on the curvelet transform, we combine the curvelet transform and the approximate flattened operator (AFO) to improve the deblended result, which can flatten seismic events approximately to preserve the details of useful signals. This is the first time that the AFO and the curvelet transform are combined to enhance the effect of the coherent-promoting operator and improve the performance of deblending. To display the advantages of the improved method, we use both simulated synthetic data and field data examples to compare and analyse the deblended results using our method and the conventional method, and confirm that the improved method can perform better.