An Efficient Recognition and Data Extraction Method for Table-Form Documents

In Asia, many documents processed in offices are table-form documents. Hence the automatic processing of table-form documents is an important issue of the office automation research. In this paper, we propose an efficient representation method for table-form documents. The representation method is based on three types of line segments. The line segments are normalized and sorted, hence the representation method provides an unified and efficient way for learning and recognition of tableforms. Based on this representation, a table-form documents may be learned interactively with the user. After a tableform document was learned, it can then be recognized and the data in data fields be extracted.