An Image-Based Method for Marine Magnetic Signal Denoising

Magnetic field measurements in the ocean environment have a low signal-to-noise ratio, which makes noise reduction challenging. Different from the conventional time and frequency domain noise filtering technics, this paper proposes an image-based marine magnetic signal denoising method. First, the one-dimensional raw measurements in a time series are converted into a time-frequency image using the short-time Fourier transform. In the time-frequency image, the signals have distinct linear characteristics. Second, a two-dimensional Fourier transform is applied to the time-frequency image. It results in a frequency spectrum image where the energy concentrations of the signal are represented as the grayscale variation in the vertical direction. Then, the curvelet transform of this spectrum image can help find the coefficients of the signal's energy distribution in the curvelet domain. Finally, a filter can be designed by retaining the coefficients of signals in directions of interest while compressing the coefficients in other directions. The proposed method is validated by both simulation and experiments. Its superior performance is confirmed by comparing it to the Gaussian low-pass filter and wavelet transform-based denoising methods.