Published as Geophysical Journal International, 216, 1214-1232, (2019)

Deblending of simultaneous-source data using a structure-oriented space-varying median filter

Yangkang Chen, Shaohuan Zu, Yufeng Wang, and Xiaohong Chen
School of Earth Sciences
Zhejiang University
Hangzhou, Zhejiang Province, China, 310027
chenyk2016@gmail.com
Modeling and Imaging Laboratory
Earth and Planetary Sciences
University of California
Santa Cruz, CA 95064
zushaohuan@qq.com
State Key Laboratory of Petroleum Resources and Prospecting
China University of Petroleum
Fuxue Road 18th
Beijing, China, 102200
hellowangyf@163.com & chenxh@cup.edu.cn

Abstract:

In seismic data processing, the median filter is usually applied along the structural direction of seismic data in order to attenuate erratic or spike-like noise. The performance of a structure-oriented median filter highly depends on the accuracy of the estimated local slope from the noisy data. When local slope contains significant error, which is usually the case for noisy data, the structure-oriented median filter will still cause severe damages to useful energy. We propose a type of structure-oriented median filter that can effectively attenuate spike-like noise even when the local slope is not accurately estimated, which we call structure-oriented space-varying median filter. A structure-oriented space-varying median filter can adaptively squeeze and stretch the window length of the median filter when applied in the locally flattened dimension of an input seismic data in order to deal with the dipping events caused by inaccurate slope estimation. We show the key difference among different types of median filters in detail and demonstrate the principle of the structure-oriented space-varying median filter method. We apply the structure-oriented space-varying median filter method to remove the spike-like blending noise arising from the simultaneous source acquisition. Synthetic and real data examples show that structure-oriented space-varying median filter can significantly improve the signal preserving performance for curving events in the seismic data. The structure-oriented space-varying median filter can also be easily embedded into an iterative deblending procedure based on the shaping regularization framework and can help obtain much improved deblending performance.