How to segment character from image?

[breta01]

How to segment character from image?

[Breta01]

haha, sounds good. I will be working on something similar, so let me know how it goes.

[Breta01]

Hi,

the first tip, I would try to remove an unnecessary noise and the small printed text from the picture. It looks like some kind of form, maybe you will be able to subtract the printed text and line. If you get just the word it should make the recognition easier and more precise.

[sagarhukkire]

Hi
Yes you are right. Its bank form, currently I am taking ROI from form by cropping then tryting to segment using CNN. Do you have any idea to filter out printed text and line. ..any link or tutorial to follow
Thanks man, your contribution is way to great to solve many OCR problems

[Breta01]

It depends on your documents. If you have an image of the empty form, which matches the filled documents, you can subtract it from filled documents. There are different approaches to it. Search for something like background subtraction opencv, but most of these tutorials are for video.

[Breta01]

Maybe this could be useful: https://stackoverflow.com/questions/44072956/opencv-image-matching-form-photo-vs-form-template

[sagarhukkire]

@Breta01 man you are great..this is what i was looking..i will visit prague in two months ..will see you for drink haha