GPR Random noise attenuation using Savitzky-Golay filter in the dual-tree complex wavelet domain

Ground penetrating radar (GPR) data, like other geophysical methods, always has unwanted energies or noise. Noise attenuation is one of the most important steps in the processing of Geophysical data before interpretation. Different methods have been proposed to attenuate the Gaussian noises of GPR data. Each of these methods has its limitations and advantages. In this study, for the first time, Savitzky-Golay (SG) filter in the dual-tree complex wavelet domain (DTCWT) have been used to attenuate the Gaussian and non-Gaussian noises of GPR data. Synthetic data results in the presence of Gaussian noise have indicated the poor performance of the soft and non-negative Garrote thresholding methods in the complex wavelet domain in such a way that the downward and damping trend of the frequency spectrum of the thresholding methods indicating the loss of the signal in the highfrequency range. In other words, when the noise is attenuated, the signal is also lost. On the other hand, applying the SG filter has indicated that the original shape of the signal has not been restored in the synthetic and real GPR data. For further investigation, the SG filter with the same wavelet field parameters in the presence of Gaussian and non-Gaussian noise has been applied to synthetic and real GPR data. The application of the designed SG-DTCWT algorithm on the GPR data has led to more reliable results in terms of signal retention and noise attenuation.