ABSTRACT
In this paper a method of pattern recognition in digital images is discussed. It is based on the best rank-(R1, R2, …, RK) decomposition of the prototype pattern tensors which are obtained from the patterns defining a class. In the case of a single prototype, a prototype pattern tensor is proposed to be constructed from the geometrically deformed versions of the available pattern. Object recognition is accomplished by comparing distances of the features obtained by projecting the test pat-
1 INTRODUCTION
The newest information technologies lead to generation and processing of enormous amounts of data. The associated problems of data processing and information retrieval require not only the newest computer technologies but also the most efficient new algorithms for data representation and recognition of patterns [Duda_2001][Tadeusiewicz_2010]. In this paper we address the problem of pattern recognition in multi-dimensional visual signals based on the best rank-(R1, R2, …, RK) tensor approximation. Its application to pattern recognition in medical images, as well as details of software implementation, is also discussed.
