Noise Removal Based on Filtered Principal Component Reconstruction

Abstract

Principal component analysis solves the overlap between the signal and noise in spectra, however the low-order components still contain high frequency spatial noise. To tackle this problem, a denosing approach based on filtered principal component reconstruction is proposed. This method designs a low pass filter group which is based on adaptive width smoothing algorithm to remove the noise of the low-order components. And then the electromagnetic data are reconstructed by the filtered low-order components. This approach can not only remove the high frequency spatial noise of low-order components, but also remove uncorrelated noise of high-order components. The experimental results of the simulation data show that the SNR is improved by 10.96 dB and 2.52 dB relative to the traditional profile filter and principal component reconstruction, respectively. The results of conductivity-depth imaging also prove that the filtered principal component reconstruction can improve the ability to identify deep targets. Finally the conductivity-depth imaging for the survey profile data indicates that noise removal based on filtered principal component reconstruction is very effective.