MultiFont/Size Character Recognition and Document Scanning

This paper presents a simple and efficient approach for the implementation of Optical Character Recognition and translation of scanned images of printed text into machine-encoded text. It makes use of four different image analysis phases followed by image detection via multi layer back propagation neural network. This paper also describes scanning the entire document and recognizing individual characters from image irrespective of their position, size and various font styles. It deals with recognition of the symbols from English language which can be easily extended for large set of symbols including UNICODE characters; this can be easily achieved by training the network for different symbol set. The phases described in this paper are simple and easy to implement.