Abstract
dc:description.abstract<p>In this thesis, I present a optical character recognition technique based on the rule-based module and object-oriented design. A document image is first scanned, and then segmented into isolated characters by the use of projection profiles. The recognition procedures include three steps: contour extraction, strokes detection and rule-based classification. The contour extraction is to extract the outbound boundary and to perform the boundary linking and noise removal. The stroke detection is to detect the features or strokes of the outbound boundary, such as horizontal/vertical lines, left/right slash lines and left-/right-opened curves. The rule-based classification consists of two rules: character definition rules and control rules. The character definition rules define the elementary criteria of combinations of detected strokes. The control rules, represented by a hashing table, embody inferences about the order in which the character recognizers are matched. The experimental results show that the recognition rate of this system reaches to 95%.</p>
Degree
thesis:*- Name thesis:degree_name
- Master of Science in Computer and Information Science - (M.S.)
- Discipline thesis:degree_discipline
- Computer and Information Science
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Gao, Min
- Contributors dc:contributor
-
- Frank Y. Shih
Subjects
dc:subject × 6Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.njit.edu/theses/2483