2

I have done an OCR application for handwritten normal characters. For the segmentation of characters I have used "histogram profile method". That successfully works for normal English characters.

I have used horizontal projection for line segmentation and vertical projection for character segmentation.

To segment lines of cursive hand written article I can use horizontal projection as previous. But I can't use same methodology for cursive English character segmentation since they are merged each other and also slanted. Can anyone please help me with a way to segment cursive characters.

Improve this question
2
  • I can vaguely remember this.. how about if you rotate the letters so they are upright, and the baseline is diagonal? Commented Jun 24, 2013 at 9:14
  • approximatively segment one then recognize it, if it is not recognized try to tune its segmentation, and so one. Commented Sep 30, 2013 at 22:35

1 Answer 1

Reset to default
3

This is a difficult problem to solve due to the variability between writers and character shapes. One option, which has achieved up to 83% accuracy, is to analyze the ligatures (connections between characters) in the writing and draw columns on the image using those ligatures as a base point. In 2013, Procedia Computer Science proposed this approach and published their research on this particular problem: https://www.sciencedirect.com/science/article/pii/S1877050913001464 (doi:10.1016/j.procs.2013.05.013)

Another approach to try is called skeletal analysis which takes the word as a whole and matches its shape with other known word shapes and predicts the word based on the entire image.

Good luck!

Improve this answer
Sign up to request clarification or add additional context in comments.

Comments

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.