| 摘要 |
<p>PROBLEM TO BE SOLVED: To correctly recognize each item on a document image of an atypical document.SOLUTION: Plural items of a detection object arranged in a predetermined direction on a document image including the plural items divided by a frame line, are determined to item definition information as a same group, and the item definition information is stored in a storage part. Then, a horizontal rule is detected on the document image and each area surrounded by the horizontal rule is detected as a frame, and a character in the detected each frame is recognized by character recognition. Based on the character recognition result, an item name in each frame is recognized, and one of the frames whose item name is recognized is selected as an attention frame. Among the frames whose item name is recognized, with the attention frame serving as reference, the frames arranged in a predetermined direction are classified to a same group of the attention frame, and based on the item definition information, a group to which item names of the frames classified to the same group belong, is determined for deciding a position of each item name, for automatically recognizing the position of each item name on the document image.</p> |
