Seismic Coherent Noise Attenuation by Combining Generalized S Transform and Empirical Mode Decomposition

Seismic data are commonly contaminated by noise. Ground roll is one of the most important noises in onshore reflection seismic data acquisition. It is a coherent noise with a lower speed, lower frequency, and stronger amplitude than reflection seismic signals. Due to the dispersive property of the ground roll, it is distributed in fan-shaped zones at near offsets about the source. There are several methods for ground roll suppression such as 1D and 2D frequency filtering and velocity filters. In this paper, a new filtering method based on the generalized S transform and the empirical mode decomposition has been used to attenuate the ground roll coherent noise. In the proposed method, the x-t domain seismic data transformed to t-f-x domain using a generalized S transform. Then, the empirical mode decomposition is applied to the transformed data along the X-direction. The first intrinsic mode function related to high wave numbers is removed to attenuate the ground roll. The denoised seismic data can be obtained by stacking up the remaining intrinsic mode functions and transforming back into the t-x domain using an inverse S transform. The efficiency of the mentioned method has been tested on real seismic data and compared with empirical mode decomposition in the f-x domain. The obtained results show that the proposed method attenuates the ground roll better than the combination of empirical mode decomposition and Fourier transform.