Sub-band invariants of handwritten texts fuzzy fragments

Abstract

Abstract Socio-economic processes informatization is largely implemented by using various documents electronic storage, including scanned handwritten texts or their fragments images, in particular, in the form of officials’ original signatures. Among the very diverse tasks of handwritten documents scans electronic repositories computer analysis, the direction of automatic search for images fragments, that contain outlines of key words of interest, is quite relevant. The search for such fragments is significantly hampered by the variability of the letterforms, even by the same author. Therefore, it is advisable to use a fragment from the analyzed text as an initial sample. Thus, we are talking about the fragments precedent identification. In this case, a complicating factor is the impossibility of using many identical objects in training, and only one is available. Therefore, the problem of forming an artificial sample, which elements preserve some common characteristic property, arises. In this work, the authors substantiated the application adequacy of the original sub-band approach, which is based on the concept of the data segment Euclidean norm sub-band part and the mathematical apparatus of sub-band matrices obtained on this basis. Decision procedures for the handwritten text fragments sub-band identification were developed, including training on a single precedent.