Improved random noise attenuation using f-x empirical mode decomposition and local similarity

Conventional f-x empirical mode decomposition (EMD) is an effective random noise attenuation method for use with seismic profiles mainly containing horizontal events. However, when a seismic event is not horizontal, the use of f-x EMD is harmful to most useful signals. Based on the framework of f-x EMD, this study proposes an improved denoising approach that retrieves lost useful signals by detecting effective signal points in a noise section using local similarity and then designing a weighting operator for retrieving signals. Compared with conventional f-x EMD, f-x predictive filtering, and f-x empirical mode decomposition predictive filtering, the new approach can preserve more useful signals and obtain a relatively cleaner denoised image. Synthetic and field data examples are shown as test performances of the proposed approach, thereby verifying the effectiveness of this method.