Aerospace Instrument-Making Annotation << Back Method of processing multispectral images of Earth remote sensing Smirnova A.I. Khachumov V.M. We consider the mathematical apparatus for processing multispectral space images and classification. The questions of processing and analysis space images, obtained from both the flow of information service video telemetry and by remote sensing (RS), remain topical and require the development of software tools, including the processing for im-ages made in different emission spectrums. Pictures taken in this way, are widely used in the space industry, both in research and development and for the solution of a large number of practical problems. Investigation of spectrums of emission and absorption, the differences of the scattering, reflection and transmission coefficients in the visible, infrared and other spectrums allow to take images, that have a number of features com-pared to conventional. To calculate distances and solving the problem of segmentation we use a metric Euclidean-Mahalanobis, which allows of better account the distribution of elements in a data set. The examples of the classification and formation of regions by the method of «filling» are given. A large amount of graphical information requires considerable computing power for its processing. The most effective in this case is the method of parallel and pipeline processing of remote sensing data on clustered systems. The scalability of computation time on a clustered installation of the domestic family «SKIF» is shown. Key words: multispectral image, processing, filling, Mahalanobis distance, scalability, program system. Contacts: E-mail: vmh48@mail.ru Pp. 50-56.