An algorithm for automatic image classification based on a bilateral criterion for homogeneity of multidimensional samples with the use of the Kolmogorov-Smirnov distance has been proposed to solve problems of thematic interpretation of aerospace video data obtained from observations of the Earth's underlying surface and cloudiness. Applicability to a set of adjusted images in multispectral channels of visible, IR, and microwave ranges of electromagnetic spectrum is a salient feature of the constructed algorithm.