The echo signal of light detection and ranging (LiDAR) is nonlinear and non-stationary and is easily disturbed by various noises. In order to filter out noises and extract effective signal information, it is necessary to select appropriate methods for noise reduction processing. In this study, Poisson noise was added to the simulated LiDAR echo signal, and then de-noising experiments were carried out by wavelet transform (WT), empirical mode decomposition (EMD), variational mode decomposition (VMD), and their improved and combined algorithms. Afterward, we selected the optimal de-noising method for LiDAR echo signal through comparative analysis. The experimental results showed that the WT-VMD joint algorithm has the maximum output signal-to-noise ratio (SNR) and the minimum root-mean-square error (RMSE) under different original SNRs, with a small smoothness of the de-noised curve, and therefore it can restore the original LiDAR echo signal well and improve the accuracy of subsequent signal inversion.