Abstract: Seismic exploration is the main method in oil and gas exploration. It becomes more and more difficult to identify the collected seismic signals, because of a sharp drop in resolution and the Signal to Noise Ratio (SNR). Conventional seismic record processing methods have shown no adaptability to very low SNR. In order to realize the requirement of high SNR, high resolution and high fidelity of seismic data processing in strong noise, this paper proposes a novel method of seismic data noise suppression based on similarity matched Wiener filtering. In this method, based on the local and non-local similarity of seismic events, the whole record is divided into overlapping sub-blocks and the blocks containing similar signals are found out by some distance measures. Then, the low rank matrix is constructed by Singular Value Decomposition (SVD) and the smaller singular values are removed representing noise. The noise in the constructed signal is effectively suppressed. Using this estimation, the Wiener filter is able to obtain a more accurate transfer function. At the same time, the effect of prior information on noise suppression is eliminated. Experiment results illustrate that the proposed method improves the SNR and effectively retains the signal amplitude.

Key words: information processing technology, Wiener filtering, block matching, singular value decomposition