摘要 |
PROBLEM TO BE SOLVED: To provide a high-precision form recognition technique for a reading object having a various kinds of forms, a type determination technique, and a means for extracting an underline shown in the form. SOLUTION: Ruled line frames 204 and 206 and a character line 212 are extracted from a form image 200, and an error of character recognition is corrected by checking a character recognition result against a word dictionary. The type of the form is determined from a feature of a table and a form name and an item name obtained by the checking. The character line and the ruled lines are extracted from the form image, the ruled lines constituting the frame is removed from the extracted ruled lines, and an underline is extracted by comparing the remaining ruled lines with the arrangement of the character line. The type of the form can accurately be determined for an atypical form such as a registered notice, and the underline can accurately be extracted without mistaking it for a stroke or the like in a character. COPYRIGHT: (C)2008,JPO&INPIT
|