Novel Approach to Segmentation of Handwritten Devnagari Word

This paper makes an attempt to segment the handwritten Devnagari words. Segmentation of script is essential for handwritten script recognition. Segmentation affects recognition so accurate segmentation is important for implementing OCR. Little work had been reported towards segmentation of handwritten text. Segmentation of handwritten words is a bit complicated as the shape of the handwritten characters is uncertain due to variability in writing styles. The proposed system carries out segmentation in hierarchical order. The system deploys the morphological operations of image processing for segmentation. Neighbourhood tracing algorithm is used for finding the segmented objects in the specific zones that correspond to constituent symbols of the Devnagari script. Segmentation accuracy is found to be 57% for segmentation of top modifiers and 55% for lower modifiers and 52% for characters in core zone.

I. Introduction

In optical character recognition (OCR), a perfect segmentation of characters is required before individual characters are recognized. An OCR has variety of commercial and physical applications. It can be used for automatic reading and processing of the forms, old degraded documents, bank cheques. It can prove as an aid for visually handicapped persons. There are many scripts and languages in India but not much research is done for recognition of handwritten Indian characters. Segmentation is a technique which partitions handwritten Hindi words into individual characters. Since recognition heavily relies on isolated characters, segmentation is a critical step for character recognition because better is the segmentation, lesser is the ambiguity encountered in recognition of candidate characters of word pieces. Casey and Lecolinet[1] describes three strategies for segmentation namely •

Classical Approach

•

Recognition based segmentation approach

•

Holistic method