“i” — A novel algorithm for optical character recognition (OCR)

Abstract

Computer vision, artificial intelligence and pattern recognition have been important areas of research for a while in the history of electronics and image processing. Optical character recognition (OCR) is one of the main aspects of computer vision and has evolved greatly since its inception. OCR is a method in which readable characters are recognized from optical data obtained digitally. Many methodologies and algorithms have been developed for this purpose using different approaches. Here we present one such approach for OCR named “ i ”. Amongst all other OCR systems available, “ i ” aims at a high speed, simple, font independent and size independent OCR system based on a unique segment extraction technique. This algorithm can be used as a kernel for single alphabet detection within a complete OCR solution system without the need for any complex mathematical operations. The highlight of this methodology is that, it does not use any libraries or databases of image matrices to recognize alphabets, but it has a unique algorithm to recognize alphabets instead. This algorithm has been implemented in MATLAB 7.14.0.739 build R2012a on a test set of 500 images of text and an accuracy of 100% for three font families namely Arial, Times New Roman and Courier New has been obtained.