Proper Names Extraction from Fax Images Combining Textual and Image Features
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 1
| Hi-index | 0.00 |
We present here the fax processing task of a messaging system called Majordomo which serves as a telephone attendant. Key information is searched in fax front covers such as names, addresses, fax/phone numbers in order to sort messages and extract condensed information. First a document analysis stage searches for physical entities at the word level. Then physical entities are discriminated into two classes : handwritten or printed. Then user specified fields identified by keywords are searched in the electronic version (Ocr'd transcription). Then fields are associated with their corresponding content if printed, with the corresponding image blocks if handwritten. The process was tested on a database of 55 fax images.