This paper focuses on suppressing random seismic noise while preserving signals and edges. We propose an edge-preserving polynomial fitting (EPPF) method leading to good signal and edge preservation. The EPPF method assumes that a 1D signal can be modeled by a polynomial. A series of shifted windows are used to estimate any sample in a 1D signal. After that, the window with the minimum fitting error is selected and its output is assigned as the final estimate for this sample. For a point in 2D seismic data, several 1D signals are extracted along different directions first and then are processed by the EPPF method. After that, we select the direction with a minimum fitting error and assign its output as the final estimate for this point. Applications with synthetic and real data sets show that the EPPF method suppresses the random seismic noise effectively while preserving the signals and edges. Comparisons of results obtained by the EPPF method, the edge-preserving smoothing (EPS) method, and the polynomial fitting (PF) method show that the EPPF method outperforms EPS and PF methods in these tests.
High-Order Polynomial Fitting Assistance for Fast Double-Peak Finding in Brillouin-Distributed Sensing
