Application of dictionary learning in marine CSEM denoising

Abstract

Marine controlled source electromagnetic (CSEM) method is widely used for submarine resources detection. The amplitudes of marine CSEM signals decay exponentially with the augment of offset. When the offset is large, the signals are easily contaminated by various kinds of noises, such as internal noises, dipole vibration noises, sea water motions noises and so on. Suppressing the influence of noises is the key to improve data accuracy. Nowadays, most denoising methods aim at removing single type noises, such as Gaussian white noises. Sparse representation provides a new way to suppress all kinds of noises once for all. Under the framework of sparse representation, the denoising effect is closely related to the chosen dictionary, a set of base functions. In general, the stronger the correlation between signal and dictionary is, the better denoising effect will reach. As a result, a new method based on dictionary learning is proposed to marine CSEM denoising procedure. Firstly, the segments suffering little from noises are captured as training set. Then the learned dictionary is constructed from the training set via K-singular value decomposition (K-SVD) algorithm. Lastly, the learned dictionary is applied to the denoising procedure by orthogonal matching pursuit (OMP) algorithm. To evaluate the effectiveness of the proposed denoising method, synthetic data examples are conducted including windowed-Fourier-transform (WFT) and wavelet-transform (WT) denoising methods, and three dictionaries (discrete-sine-transform (DST) dictionary, DST-wavelet merged dictionary and the learned dictionary) under sparse representation framework. The comparisons demonstrate the improvement of the proposed dictionary-learning-based denoising method. Note: This paper was accepted into the Technical Program but was not presented at IMAGE 2021 in Denver, Colorado.