The display of information characterizing the distribution of thermal gradients is associated with the formation of a twodimensional image. A thermal image is most often a group of objects with varying intensities. This picture is due to physical aspects. Most often, the thermal background does not have a clear boundary, and therefore the use of conventional methods for detecting special features in the form of angles or borders is not applicable. In the previously described works, a method was proposed for identifying special features based on the search for the midline, for closed objects. This approach did not take into account the possibility of changing the degree of growth of the line displacement, which may be correct provided that a pair of objects of different temperatures adjoin each other and there is information about this. The article proposes to develop a modification of the search method an equidistant curve that allows you to introduce a curve growth function. The difference from the original approach is the possibility of regulating the degree of movement of the internal or external borders in the direction of the neighbor. The approach proposed in the work includes the operations of filtering, changing the color space, changing the color range, and searching for stable signs. The task is relevant and may also be applicable in tasks: increasing the clarity of x-ray images; eliminate blur in photos; preprocessing images obtained in the visible spectrum; deblurring images obtained in motion, etc. In the article on a set of test data shows the applicability of the proposed approach for identifying stable features in thermal imaging images. The data obtained by FLIR C2 and SEEK thermal imaging cameras cover a wide range of possible applications from analysis of data obtained outside buildings, as well as technical processes occurring inside mechanisms.