Fuzzy approach to document recognition

The authors present a new approach to document recognition using fuzzy rules. The system uses information such as relative locations, relative sizes, and positions. A prototype DOCREC-III is described, which takes bitmap scanned images as input and uses spatial knowledge (layout structure) to reason about the rectangular segments (logical structure) in technical papers. The system provides a compact rule base with accuracy and efficiency. The rules are concise but powerful enough to recognize a variety of layout structures observed in technical papers.<
>