| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To precisely, efficiently detect an image area including a specific character string. <P>SOLUTION: When a rectangle candidate is detected, OCR processing in the rectangle is tried in a step S203. In a step S204, knowledge processing is executed by using the acquired result of the OCR processing. The knowledge processing is the processing for making amendment processing to realize the recognition result of a word 'Iwate' in case that there is the word 'Iwate' in a recognition object and the word is registered with a knowledge dictionary (character recognition dictionary). In a step S205, it is determined what has become of the result of the knowledge processing using the character recognition dictionary in the step S204. When the knowledge processing is successful and a land transportation branch name is present, even if an amendment is not made, or an amendment to the land transportation branch name is possible, the rectangle is determined to be an image rectangle to be found. Then, in a step S206, the processing that reflects the result of the knowledge processing in the step S204 is executed, and then the detection processing is finished. <P>COPYRIGHT: (C)2009,JPO&INPIT</p> |
