A suitable denoising method is explored to reduce the influence of noise on atmospheric Rayleigh lidar echo signal. Combined with simulated and measured echo signals, the denoising effects of moving average, Hanning window sliding-window, wavelet transform (WT), ensemble empirical mode decomposition (EEMD), WT-EEMD-LOWESS, and EEMD-VMD-IMWOA methods are compared with respect to their influence on photon number profile, atmospheric fluctuation information, and temperature retrieval accuracy. The analysis results show that the temperature retrieval results processed by the Hanning window sliding-window and the EEMD-VMD-IMWOA methods have high accuracy and effectively retain the atmospheric fluctuation information, surpassing the other methods. However, EEMD-VMD-IMWOA method has better robustness, and the Hanning window sliding-window method is not suitable for processing high signal-to-noise ratio signals. If the signal-to-noise ratio of the measured signal is in the range of 0?200, the Hanning window sliding-window method can be selected for denoising; otherwise, EEMD-VMD-IMWOA method should be preferred to improve retrieval accuracy.