An attempt to perform Bengali optical character recognition
Persi W. Diaconis, Arnab Chakraborty · 2003
This thesis provides details about an attempt to perform optical character recognition (OCR) for the Bengali language, used by many people in north east India. In particular, a new algorithm is proposed. After describing the necessity of performing OCR for this language, the thesis makes a selective review of the literature relevant to OCR. Then details are furnished about various aspects of the new algorithm. These include image segmentation, glyph recognition using a sequential learning algorithm, and postprocessing the output. Pointers are provided to an implementation of the algorithm in Java.