Model-based analysis of printed tables

We develop a strategy for extracting the underlying relational information from the images of printed tables. Visual clues that exist in the image are used for extracting first the physical, and then the logical structure of the table. Since these visual clues generally have a logical meaning, there must be some association made between the graphical attributes extracted and their function of reflecting the logic expressed by the table; this knowledge is coordinated in a model. This approach, therefore, can be adapted to all tables which have graphical attributes discernible to the image analysis being used