ABSTRACT
We present a method for extracting structural information of a table and a method for recognizing hand-printed or drafted or machine-printed text from an image of a telephone company tabular drawing. The structure interpretation procedure uses horizontal and vertical lines and white streams as demarcation cues to identify each text block in the table. The character recognition system uses statistical and neural network techniques and a-priori field grammars for interpreting text strings. We are currently working on integrating the two modules.
