Preprocessing tilted text for Tesseract OCR using opencv

Asked Mar 08 '20 at 04:44

Active Mar 10 '20 at 15:20

Viewed 407 times

I am trying to identify the words on the back of a medicine pack, This is an example of my picture. this but better quality :

Since the text is tilted, where should I start in trying to auto-rotate it such the OCR works?

I've looked at this document and this other question on Stack Overflow but in these cases the text is more than one or two words so the textlines are longer.

edited Mar 10 '20 at 15:20

asked Mar 08 '20 at 04:44

Please check your image format.and edit it. – jizhihaoSAMA Mar 08 '20 at 04:54
Sorry what do you mean? Something wrong with the image link? – cheap brent crude Mar 08 '20 at 05:08
No,But just edit it with ``markdown`` format,it can make people who want to answer your question more convenient. – jizhihaoSAMA Mar 08 '20 at 05:16
Unfortunately I can't embed images as a new poster but thanks for the suggestion – cheap brent crude Mar 08 '20 at 05:27
1

You’re going to need **Much** better quality images – DisappointedByUnaccountableMod Mar 08 '20 at 09:04
Did you actually try the methods in the pages you linked to? Hough line detection can work even with shorter lines (you control how long lines it will detect) though no doubt you would have to adapt to your specific images. – DisappointedByUnaccountableMod Mar 08 '20 at 09:09

0 Answers0