A denoising method combined with Hilbert-Huang transform (HHT) time-frequency analysis for backscatter signal of laser diode ceilometer based on HHT and Savitzky-Golay filtering is proposed. Considering characteristics of ceilometer backscatter signal proper structure of adaptive time-variant time-space filtering bank (TFB) assisted by Savitzky-Golay filtering is selected to suppress noise. Then, HHT time-frequency analysis is used to observe signal features and test denoising effect. It is proved by plenty of test data under various weather conditions that the method can help to suppress noise without deteriorating signal features and improve the ability of cloud base and height identification, and finally obtain precise retrieving of vertical visibility.