In this paper we propose a new method of filtering Canopy Height Model (CHM) with accuracy control (threshold filter) based on the Gaussian filter in GIS. Standard methods of image filtering have a significant drawback – then the size of the moving window or number of iterations increased, along with the elimination of noise, there is the uncontrolled loss of accuracy. The solution to this problem must be sought in the filtering process, which allows control of the degree of smoothing, for example, using a threshold value. Selection of the threshold value depends on the required accuracy, scale and detail. In general, using the number of N iterations, it is possible to smooth the CHM within a preset threshold.