Introduction. Possible methods of pattern recognition are described. Transform coefficient, allowing numerically to evaluate the difference between the test and reference signal is proposed. In this case, the reference signal is selected by researcher independently, which gives more freedom. Theoretical positions. The possibility ofMellin transform using for pattern classification of images based on their normalization or normalized transformation when scale of studied images is differ from a reference image, is considered. A similar ofFourier and Mellin transforms is proved. Classification algorithm. Proposed algorithm can be used when change of scale arguments of the signals is presented. This algorithm has a clear structure that allows implement it in hardware with minimal effort. Examples. Examples of the Mellin transform using for different distortion types of the test signal are considered. The test signals have time scale and distorted form changing. The obtained sensitivity value of classifier to parameters changes of the investigated image with respect to the reference image was sufficient to get a stable work of this unit. Conclusions. The main advantages of the Mellin transform using for recognizing signals at different scales are presented is conclusions.

Показана возможность использования преобразования Меллина при классификации образов на базе их нормализации или нормального преобразования при изменении масштаба аргументов исследуемых образов относительно эталонного образа. Приведен пример вычислений, показавший простоту применения преобразования Меллина. Получена достаточно высокая чувствительность классификатора к изменениям параметров исследуемого образа относительно эталона.