A method based on the wavelet threshold denoising algorithm is proposed for the suppression of solar background light, so that the separation of the real signal from the noise in the Raman returned signal can be realized and the background noise in daytime can be removed. Based on all-day data measured by atmosphere water vapor Raman lidar system built in Xi’an University of Technology, influences of decomposition level, wavelet function, threshold function, and threshold selection method on the denoising results of returned signal in daytime are discussed. Signals before and after denoising are compared and denoising evaluation functions are compared. We adopt wavelet sym6, decomposition of five layers, improved threshold function, and improved threshold method to obtain the better denoising effect for water vapor Raman and Mie-Rayleigh scattering signals in daytime. Furthermore, profiles of the atmospheric water vapor mixing ratio, and the results of signal-to-noise ratio (SNR) of water vapor are discussed. Results show that SNR for lidar water vapor measurement increases by 3.4 times in the denoising process. and the water vapor detection range can be improved up to over 3 km from 1.5-2 km in daytime. Lidar continuous detection experiments and denosing process are carried out during 24 h. Variation characteristics of the atmospheric water vapor mixing ratio are obtained below boundary layer, and the results agree with data from near-surface weather stations. It is verified the feasibility and effectiveness of the wavelet denoising algorithm used in all-day atmospheric water vapor detection.