AN IMPROVED ROBUST THRESHOLD FOR VARIATIONAL MODE DECOMPOSITION BASED DENOISING IN THE FREQUENCY-OFFSET DOMAIN

摘要

We proposed a novel robust denoising method using variational-mode decomposition (VMD) and the detrended fluctuation analysis (DFA) in the frequency-offset (f-x) domain, named robust DFA-VMD. DFA is mainly introduced to solve the problem that VMD requires the number of modes to be predefined. The scaling exponent obtained by DFA is a robust metric to measure the long-range correlations and can be used to adjust the number of intrinsic mode functions (IMFs) automatically. To reconstruct the denoised signal, a scaling exponent is also used as a threshold to identify and remove the noisy modes. We define a novel robust threshold of random noise in seismic data, because the predefined noise boundaries for other time series cannot perform perfectly when dealing with seismic data. The proposed robust DFA-VMD is an almost parameters-free denoising approach and we apply it in the (f-x) domain for seismic denoising. We have verified its performance by comparing it with the results from several other methods including (f-x) deconvolution and the conventional DFA-VMD. Two synthetic examples and three field-data examples revealed the effectiveness of the proposed approach in applications to random and coherent noise attenuation.