How can I OCR cursive text?

Rubanraj11 June 17, 2014
Pinterest Stumbleupon Whatsapp

How to convert a cursive word image into word document or text ? Kindly tell me with clear steps please ….I had tried OCR but it fails to produce the contents.

  1. Hovsep A
    June 17, 2014 at 11:56 am
  2. Oron Joffe
    June 17, 2014 at 9:56 am

    If by "cursive" you mean hand writing, there's no general answer to that. There are specialised systems for recognising handwriting (sometimes called ICR "Intelligent Character Recognition"), but they are only accurate when used in very specific domains such as recognising addresses, where they can cross-check the recognised text against a database of known addresses.

    Also, just to clarify, it's the conversion of scanned handwriting which is so difficult. Recognising handwriting in action (e.g. on a tablet) is much easier and more accurate, as the movement of the stylus provides important extra information. You'll find more information on the subject at http://en.wikipedia.org/wiki/Handwriting_recognition

    If you meant something else, please clarify.

    • Rubanraj11
      June 17, 2014 at 4:01 pm

      It is not a hand written one, its a scanned image file that contains cursive letters (fonts)
      Kindly help me

    • Oron J
      June 17, 2014 at 11:00 pm

      It would be useful to have an idea of what the font actually looks like... Cursive can mean different things to different people. Basically, if the font is reasonably simple but slanted (similar to ordinary italics), you can use a good OCR package such as Nuance OmniPage or Abbyy FineReader Professional and it will read the text well. The more ornate the font, with swishes, decorations and so on, the less accurate the recognition will be. If the letters are joined to each other (the proper definition of "cursive") then no OCR package will read it, as it is essentially like handwriting, but less important commercially (i.e. very little material is printed in properly cursive text).

    • Rubanraj11
      June 18, 2014 at 1:54 am

      [IMG]http://i60.tinypic.com/5bwc5f.gif[/IMG] I have attached the file take a look at it Oron at suggest the best method to convert it into word doc

    • Oron J
      June 18, 2014 at 10:22 am

      I'm afraid this is just too cursive for OCR software. I've just tried it in Abbyy FineReader 11 Pro - the best OCR I've ever used (there's now a version 12 but I haven't got it), and the first few lines came out as follows:

      iitnxzctinp, it, tin, at thii, moment, ampuazf.' “ImiApe to enpineminp, "called, kinh, itill quiethf,. “tfei, tin? imneachf,, i^that'iwhattfounewondeeinp,,' antioened icott'i voice. “welt . . . 'kinh, hetitated. vejun,wai,neadtf, to take them, to itiventf-knain, cone. wai, it potiihle that thene could he

      The results may improve somwhat with a higher resolution image. If you like, scan the image again at 300 Dots Per Inch and let me know and I'll run it through FineReader again and post the results.

    • Bruce E
      June 18, 2014 at 5:51 pm

      If you rescan the image, do it at the highest resolution your scanner supports and save it as a TIFF file (lossless format as opposed to the lossy JPEG format).

Ads by Google