Devanagari Characters Recognition: Extracting Best Match for Photographed Text

Devanagari script is used by most of the people in India. There are some scriptspecific structural characteristics of Devanagari script which makes the character recognition problem more challenging. Many OCR tools are available for printed or handwritten Devanagari script recognition. In these systems the input is given in the form of images of the script which can be scanned or photographed. But the existing systems are not robust. They give unexpected results when the input to the system is not ideal that is the image is rotated or tilted or has illumination variance. Our goal is to build a robust OCR system for printed Marathi highlighted text where the variation with respect to font, size, orientation and illumination are allowed. This paper proposes appropriate image transformation techniques to get a robust Devanagari Character Recognition System