Poor result with horizontal lines in text

[Moritz Leiss]

Hi folks,

i've been playing a while with tesseract and opencv to get the best out of my scans. 

But lately i came across this problem:

I need to scan a bunch of documents which are printed by an old needle-printer(I suppose), which

has a thin "no-ink"-line horizontally through the text (s. attached pichture).

With these documents i get no or very poor results. 

Could some one point me in the right direction how to get tesseract to read them? Is there some

image-preprocessing I could do? Or do I have to train tesseract this "broken font"? (...that would

be bad, because this line is not alway at the same position within the font).

All help welcome :)

Thanks,

Mo

[Moritz Leiss]

Sorry, forgot the Attachmet.. here it is :)

Btw. the output is:
"E...

A_.

1..

O

1..

2

2

1..

0

A1

1..

A1

A1

1..

8

O

0

5

no

:2.

8

5

"