Off-line Cursive Handwritten Word Segmentation, A new approach

The segmentation of off-line cursive handwritten word is an important step in cursive handwriting recognition. In this paper a new, simple yet effective approach is proposed. Proposed technique is based on the analysis of the ligatures of the characters in the cursive word. The only preprocessing is to skeleton the word to allow variations in pen thickness and tilt in writing. There is no constraining for the writer. To make a fair comparison benchmark database IAM is used for experiments. Results obtained thus show that the proposed approach is capable of accurately locating the letter boundaries in unconstraint cursive handwritten words.