Abstract: In-seam wave recording typically has a lower signal-to-noise ratio,and a dispersive characteristic that makes the conventional denoising methods difficult to be applied to in-seam wave data. Therefore,the polarization filtering in frequency domain is used to suppress the noise of in-seam wave recording based on the difference between in-seam wave and interference wave in polarization direction and polarization degree. Furthermore,the authors improve the po- larization filtering weight function in frequency domain. The polarization filtering uses multi-tapers spectral estimation to obtain the polarization coefficients of the in-seam wave and the interference wave,and designs polarization filter to suppress the noise in frequency domain. By improving the weight function of the polarization filtering,the transition of coefficient between the noise and the effective signal is reduced. Accordingly,the signal-to-noise ratio of in-seam wave data with strong background noise has been greatly improved by the polarization filtering. In addition,the authors elim- inate the influence of noise on the effective signal before the polarization coefficient is obtained. It is achieved by cal- culating the spectral density matrix of noise according to the background noise before first arrival,so that the polariza- tion coefficient of the effective signal is more accurate. The actual three component in-seam wave data processing re- sults show that this method has obvious denoising effect on the channel record with low signal-to-noise ratio.