According to the characteristic of lidar-return noise, an effective de-noising method of wavelet packet based on average threshold is presented and described to reduce the lidar-return noise. In the method, the thresholds of every node of the best wavelet packet basis are acquired and averaged, and then the average value is used as a global threshold to quantize the decomposition coefficient of every node. To verify the feasibility of the de-noising method, some numerical simulations are carried out to reduce the Gauss white noise of the simulated signal, and also, the Mie lidar returns contaminated by noise are de-noised by the method. Moreover, the comparisons among the wavelet packet transform based on average threshold, wavelet transform based on global threshold and the wavelet packet transform based on default threshold are performed. Experiment results show that the method of wavelet packet based on average threshold is capable of effectively reducing the lidar return noise and the de-noising effect is better than other two methods.