Character recognition with variable subdivisions of a character region

摘要: An input, unknown character is optionally read to form a binary image, and plurality of directionality codes are assigned pixels defining the contour image selectively define code-assigned which then subdivided into sub-regions in mesh such that each sub-region includes substantially equal number codes. Then, histogram defined for collection histograms input character. The compared with registered, known characters, having similar histograms, one greatest similarity selected identify determined by calculating distance between registered characters. Preferably, subjected line-broadening processing prior step assigning