The new algorithm for handwritten characters recognition based on the digital image conversion into planar graph

Abstract

Abstract—The paper presents the new algorithm for handwritten characters recognition. The algorithm is based on the procedure for converting an image into a planar graph, where the lines in the image correspond to paths in a flat graph, that is, a sequence of graph vertices connected by edges. A special function is used to determine the shape of a simple smooth curve without self-intersections. This function is invariant with respect to such transformations of the plane as rotation and scaling, which makes it very convenient for recognizing handwritten characters. The mathematical construction of the function of the angular characteristics of the curve can easily be transferred to simple broken lines. The window vector Fourier transform coefficients are used as feature vectors to identify polylines corresponding to paths in a flat graph.

Abstract—The paper presents the new algorithm for handwritten characters recognition. The algorithm is based on the procedure for converting an image into a planar graph, where the lines in the image correspond to paths in a flat graph, that is, a sequence of graph vertices connected by edges. A special function is used to determine the shape of a simple smooth curve without self-intersections. This function is invariant with respect to such transformations of the plane as rotation and scaling, which makes it very convenient for recognizing handwritten characters. The mathematical construction of the function of the angular characteristics of the curve can easily be transferred to simple broken lines. The window vector Fourier transform coefficients are used as feature vectors to identify polylines corresponding to paths in a flat graph.