Seismic edge-preserving smoothing based on Morlet wavelet transform

Seismic edge-preserving smoothing (EPS) can suppress noise while preserving edges of important geologic natures. Edge detection is the key prerequisite for EPS. Edges are usually estimated from seismic data with full frequency, but noise often contaminates such estimate. In a recently-published paper, edges are estimated from the dominant frequency data because they have high Signal Noise Ratio (SNR), but it will bring a mismatch between the estimated edges and the real edges of different frequency data. The seismic data of different frequency has different SNR. The higher SNR data can provide higher accuracy of edge estimating. Morlet wavelet has good local performance and is suitable for decomposing non-stationary seismic signal, so it is selected to divide seismic data into several division frequency data based on multiple frequency. For the seismic data with not too low SNR, at least 2~3 division frequency data have higher SNR. When these high-SNR division frequency data are scanned, their edge information can be obtained. Then the division frequency data of that frequency and similar frequency are smoothed on the basis of the acquired edge information. To ensure the stability of the filtering algorithm, different smoothing methods are used according to whether there is an edge. The ultimate noise-removed seismic data is the combination of the smoothed division data with different frequencies. Application to synthetic model and field data shows the method can effectively reduce noise, preserve edges and preserve trackable events.