Application of topology to optical recognition

The handwritten symbol is formed from a small number of thin lines. When scanning its image, a Boolean matrix appears, the units of which correspond to the pixels of the image, and the zeros correspond to the pixels of the background. The thickness of the lines that make up the symbol image depends on the resolution of the scanner and can be very significant. The report will discuss topology-based algorithms that make it possible to find a subset of thin lines in the Boolean matrix of a symbol, preserving their essential characteristics for character recognition, such as the direction, relative length, type of bulges, and intersection structure.