JPS6419487A - Character recognizing device - Google Patents

Abstract

PURPOSE:To easily recognize a document containing the information being not a character of a graphic and a photograph, etc., by instructing only a character part to be recognized, as a recognition area by using a pointing device with respect to a recognition object image which is displayed on a display part, and recognizing an image data corresponding to the instructed area. CONSTITUTION:A recognition area instructing part 4 instructs an area to be recognized, by a pointing device with respect to an image which is displayed by a display part 3, and a character segmenting part 5 reads out a recognition object image corresponding to the area instructed by the recognition area instruction part 4 from an image memory 2, and segments each character pattern contained in the recognition object image by a rectangle. A feature extracting part 6 derives the number of strokes of the character pattern segmented by the character segmenting part 5 as a feature quantity, and a classifying part 7 derives a distance of the feature quantity of the character pattern derived by the feature extracting part 6, and the feature quantity of each character registered in advance in a dictionary 8 and sets a character of the smallest distance as a recognition candidate character. In such a way, a document containing the information being not a character, of a graphic and a photograph, etc., can be recognized without segmenting the information being not a character.