OCR system for poor quality images using chain-code representation

Abstract

The field of Optical Character Recognition (OCR) has gained more attention in the recent years because of its importance and applications. Some examples of OCR are: video indexing, references archiving, car-plate recognition, and data entry. In this work a robust system for OCR is presented. The proposed system recognizes text in poor quality images. Characters are extracted from the given poor quality image to be recognized using chain-code representation. The proposed system uses Google online spelling to suggest replacements for words which are misspelled during the recognition process. For evaluating the proposed system, the born-digital dataset ICDAR is used. The proposed system achieves 74.02 % correctly recognized word rate. The results demonstrate that the proposed system recognizes text in poor quality images efficiently.