Based on the theory of multiple-reference adaptive noise cancellation and wavelet transform modulus maxima, this paper studied the method of interference cancellation for vertical deformation data from Dushanzi. The results indicate that: The annual variation in vertical deformation data is produced by seasonal variety of near-surface temperature and the temperature difference between fault′s hanging wall and foot wall, and the data is also influenced by white noise and other random noise. By using the data of meteorological temperature and auxiliary temperature to carry out adaptive noise cancellation, we can eliminate the effect of annual variation. Furthermore, by means of wavelet transform modulus maxima denoising on the data which has been ruled out the effect of annual variation, the white noise is inhibited, and the signal′s distortion is clearer and more accurate. Simulation testing and practical application demonstrate the denoise effect of the said method is effective and practical. The vertical deformation data that eliminated the effect of interference is found in good correspondence to the earthquakes of Ms≥4 within the radius of 100km from the station, and that reflects the station′s ability to response earthquake of Ms≥4 for surrounding area.