Tesseract misreading numbers

[ilochray]

I am using Tesseract 3.0.5 with the .Net wrapper installed from the Nuget package https://www.nuget.org/packages/Tesseract/ .  I have a program which reads text from an image.  I am using PSM = 6 and pre-process the image to remove noise.  I select a portion of the image for processing and have saved the image portion using the tessedit_write_images variable.  I attach the image.

My problem is that the character "6" in this image is always read as "5".  In my program, I iterate through Words.  In each word that should contain a "6", it is read as a "5".

I am using the standard tessdata files.

If I use the command line tesseract.exe to read the file it reads the file 100% correctly.

Why is the API behaving differently and mis-reading the numbers?

[Lorenzo Bolzani]

Is the image you attached the one saved from the API with the tessedit_write_images options? Or is the one you give as input to your program?

If it is not saved from the API please try to save the image as PNG immediately before the API call and compare it to the input one.

Try to use a single line, PSM 7 and see if you have the same problem. Just trying to better understand the problem.

I do not know the Nuget wrapper, maybe it is passing some extra options by default, but it is unlikely. With python tesserocr I get the same results from API and command line.

I set lang, oem mode and psm mode when I use the API.

Bye

Lorenzo

--
You received this message because you are subscribed to the Google Groups "tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tesseract-oc...@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/ff7f4027-c18b-4f74-beb6-d60aa83481a5%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.