Efficient algorithms for segmentation and recognition of printed characters in document processing
Shaofeng Liang, M. Shridhar, Majid A. Ahmadi · 2002
A novel discrimination function for segmenting touching characters based on both pixel projection and profile projection is given. A dynamic recursive segmentation algorithm is developed for effectively segmenting touching characters. Contextual information is used to correct errors. An effective method for detecting and correcting the slant in italicized fonts is also presented. The overall character recognition rate is 99.85% based on the total of 36,000 characters with about 40% touching characters.>