The purpose of the article is to develop and study the method of noise removal on low spatial resolution thermograms obtained by commercial small-sized thermal imaging cameras, which due to their weight and sizes can be used on mini-UAVs for thermal monitoring of power facilities. The main parameter that carries information about the operation mode of the object and the presence of defects during thermal monitoring is the temperature distribution over the surface of the object, which can be recorded by a thermal camera. The characteristics of commercial thermal imagers with uncooled arrays were analyzed. Due to low cost and mass, their usage can significantly reduce the cost of the monitoring system by reducing both the cost of the payload and the requirements for UAV payload capabilities. An issue of thermal imaging is a relatively low resolution of thermograms and a significant level of noise on them, which requires the image processing methods that would minimize the noise level. The process of obtaining thermograms using a compact thermal imager in real-time was considered. A method of removing the main types of thermogram distortions, based on an adaptive digital median filter, was proposed and implemented. The process of occurrence of a thermal gradient that occurs during the operation of the bolometric matrix was analyzed, and a method for was removal is given. It is shown that the proposed methods allow almost completely eliminated the thermogram artifacts.