help with OCR

[Ehlert]

I am doing a project where I need to do OCR on some printed lables

the font size is on 4...

and I am having a hard time seperating the letters - I am getting closer, but any hints would be highly appriciated!

ocr1 - orginal image

ocr3 - incresed brightness and contrasts

ocr4 - converted to binary, and used erode

[mfitzsimons]

Try Gray Morphology - Dilate followed by Threshold.  To separate characters you need to be able to draw a straight line vertically between characters without touching anything (blank space).  You need to have at least one-pixel between characters.  See attached image and example.

 

[Ehlert]

Thanks a lot, will try this tomorrow
Why dialate and not erode?

[mfitzsimons]

Yes, seems to work in reverse of what you would think.  Play with both and you will see why I chose dilate.

Matt

[Ehlert]

I will 🙂
And I was not aware that a straight horizontal line was needed to separate the letters, that explains a lot!

[Ehlert]

Meant Vertical line of course ...

[mfitzsimons]

That is why you get to enter a security phrase when you buy tickets to prevent OCR automated buying.