Simultaneous denoising and preserving of seismic signals by multiscale time-frequency peak filtering

• Time-frequency peak filtering (TFPF) is an effective method to suppress random noise.
• TFPF adjusts the filtering window length adaptively by scale decomposition.
• Multiscale TFPF solves the contradiction in window length selection.
• Multiscale TFPF makes a tradeoff between signal preservation and noise attenuation.

Time frequency peak filtering has been successfully applied to eliminate pervasive random noise in the time-frequency domain. The linearity of the signal is crucial for denoising in the time frequency peak filtering method. We usually apply pseudo Wigner–Ville distribution to make the signal locally linear in time. However, there is a pair of contradiction in window length selection for pseudo Wigner–Ville distribution. If we choose a short window length for pseudo Wigner–Ville distribution in the time frequency peak filtering, it leads to good preservation for signals, but the denoising performance is relatively poor. So the contradiction between the signal preservation and noise attenuation cannot be solved by a fixed window length. In this paper, we present a multiscale time frequency peak filtering to solve this problem. In the novel method, we adopt a Laplacian pyramid to decompose the seismic data into multiple scale components. These components have different frequencies. Then a short window length can be chosen for signal-dominant scale to preserve the signal and a long window length is applied to noise-dominant scale by the time frequency peak filtering to suppress more noise. We test the performance of our proposed method on both synthetic and real seismic data. Tests demonstrate that the multiscale time frequency peak filtering based on Laplacian pyramid can eliminate the random noise more effectively and preserve events of interest better than the conventional time frequency peak filtering.

Figure optionsDownload as PowerPoint slide