ICESat-2 is the world's first laser altimetry satellite to use photon counting technology. It provides high-precision， three-dimensional， large-scale ground data that can be obtained rapidly. Altimetry data obtained using photon detection systems generally contain ground signals and background signals， such as atmospheric scattering， which must be filtered to obtain ground information. To analyze the distribution characteristics of ICESat-2 background and signal photons and the performance of the point cloud filtering algorithm， we selected data from six types of land cover （city， sea ice， desert， vegetation， ocean and ice sheet （or glacier）） in different observation conditions. A statistical analysis of the background rate was then performed. The results show that the average background rate of daytime observation data is 10⁶ points per second， which is much higher than that of nighttime observation data （10⁴ points per second）. The background rate of the weak beam is equivalent to that of the strong beam. Among the six types of land cover， the ice sheet （or glacier） has the highest background rate. Based on the statistical results， 21 datasets and seven representative point cloud filters were selected for denoising experiments. After analyzing the accuracy， we can conclude that the improved local density method with recall， accuracy， and F-measure values greater than 0.90 has the best denoising performance and is relatively stable. Finally， the performance， characteristics， and applicability of the seven filtering algorithms were summarized and analyzed. This can serve as a reference for the subsequent data use and selection of filtering algorithms.