SCANTAB: TABLE RECOGNITION BY REFERENCE TABLES

The ScanTab system represents a knowledge-based approach to table recognition in scanned documents. In contrast to most systems which recognize tables by grouping layout information, our system uses predefined information about which table types may appear. This enables a very accurate detection able to cope with distorted tables and tables providing little layout information, e.g., no lines, bad alignment, or few rows. Table recognition starts with the detection of the table header. Afterwards, this header is compared with table headers of known reference tables. Having determined the correct reference table, the information kept in the knowledge base is utilized to compute the complete table structure. A graphical user interface allows an easy and fast specification of reference tables.

[1]  Osamu Hori,et al.  Robust table-form structure analysis based on box-driven reasoning , 1995, Proceedings of 3rd International Conference on Document Analysis and Recognition.

[2]  Edward A. Green,et al.  Model-based analysis of printed tables , 1995, Proceedings of 3rd International Conference on Document Analysis and Recognition.

[3]  Andreas Dengel,et al.  Message extraction from printed documents-a complete solution , 1997, Proceedings of the Fourth International Conference on Document Analysis and Recognition.