Edge based directional features for English-Bengali script recognition
Anurima Tikader, Niladri Bihari Puhan · 2014
In a multi-lingual and multi-script country like India, we often have to encounter with documents which are written in different scripts. For electronic processing of documents, the characters present in that document are identified with the help of optical character recognition (OCR) algorithms. To a large extent, the performance of such recognition algorithms is script dependent, i.e. they perform well on a particular script. So in a multi-script environment, an OCR bank containing a number of algorithms for different scripts is preferred. Automatic script recognition first identifies the type of script and directs the input document to a suitable OCR scheme for further processing. In this paper, a new method for recognition of English and Bengali scripts is proposed. Edge based directional (vertical and diagonal) features are used in the proposed method. The overall script recognition accuracy achieved by the new method for printed and handwritten image databases is 99.5% and 98% respectively.