This paper describes a novel technique for optical character recognition of handwritten text using the basic geometrical strokes contained in the alphabets of a language. It is observed that all the characters of a language can be represented as a set of connected basic geometrical strokes; thus if we break a ligature, even if the ligature contains more than one character, as in the case of cursive languages, the technique can determine/recognize the characters contained in the ligature. The recognition of characters is font independent: however it is also possible to recognize the typed characters of the standard fonts by employing this technique. Hence font based character recognition is a special case of the proposed technique. The technique was implemented by developing a C#.NET application called LIOCR (language independent optical character reader). The results obtained after applying LIOCR to 25 samples of handwritten text have also been reported.
展开▼
