| 摘要 |
PROBLEM TO BE SOLVED: To enhance the accuracy of the correction of a recognition result and recognition by language analyzing the character recognition result in a character recognizing device. SOLUTION: This character recognizing device is equipped with a character segmenting part 30 for segmenting a character segmented with a line segmenting part 20 segmenting a character line from a character region of an input image; a character recognizing part 40 for obtaining a recognizing candidate from character characteristics from the character segmenting part and a recognition dictionary 45; a language analyzing part 50 for correcting the character candidate; and in addition a line type classifying part 60 for classifying lines segmented in the line segmenting part into a text line, a character decoration line (a ruby, an underline, an emphasizing side point, half-tone dot meshing, or the like); a character decoration sampling part 70 for extracting character decoration when a classification type in the line type classifying part is character decoration; and a corresponding character extracting part 80 for extracting a character row of the adjacent line corresponding to the character decoration. The language analyzing part 50 conducts form analysis to the combination of candidates outputted from the character recognizing part with a word dictionary 55, limits the range of a morpheme from the information of the character modification line, and removes the combination with no possibility as a sentence to enhance collation accuracy.
|
