Community

Search a String in an Image

Hello, I would like to know if I can use Abbyy SDK for Android to search for specific numeric string in an Image (TIF or JPG) and zoom in to that place where it was found?

0

Comments

3 comments

  • Avatar
    Oksana Serdyuk

    There are two different SDKs which you could try to use for your usage scenario:
     
    • online solution – ABBYY Cloud OCR SDK. This is a platform-independent service that provides Web API for OCR. It is implemented using the REST software architecture principles, so it can be used in any development tool supporting communication over the network. The processing is performed on ABBYY’s side, so a user do not need to use the resources of his device for OCR. Cloud OCR SDK is based on our common OCR technologies, and it is intended for processing images (scans or photos). Please feel free to follow the link and after successful registration, the technology will be accessible to test.

    • offline solution –  ABBYY Real-Time Recognition SDK. This is a software which should be used on the device, the processing is performed on the user’s side. In comparison with ordinary OCR technologies which work with scans and photos, RTR SDK has two options:
       1. The benefit of this SDK is that you can use video frames to capture information from a document. We combine several frames of video in real-time and then merge the results to give a more accurate reading of the text. The end user would simply hover their device over the document until all required text has been correctly recognized.
       2. The second option is to simply take a standard still image using a mobile device and then perform full text OCR on the entire document. Here you can receive access to full textual and layout information.

     

    You can try both variants which are free for trial use.

    0
  • Avatar
    Kekus
    Ok sounds good. The second offline option will fit perfectly. So I can get the coordinates from the obtained string, right? How small can the font size be on a static image till the ocr get unreliable? I’m talking about number Sequenz (1, 2, ... 999)
    0
  • Avatar
    Oksana Serdyuk

    Yes, you can get the coordinates of the recognized text. RTR SDK returns the recognized result as text lines with their coordinates represented by objects of the TextLine class. Each TextLine contains the information about the enclosing quadrangle for a single line of text and the recognized text as a string, as well as some additional information about the characters. Please see the details in the Developer’s Documentation: TextLine class

    Regarding the font size, for accurate recognition the images must conform to certain requirements:

    • the letters' size must be 2 to 20 mm, and not less than 10 pixels on image
    • the best resolution for texts printed in fonts 10 pt or larger is 300 dpi
    • the best resolution for texts printed in fonts 9 pt or smaller is 400-600 dpi.
    0

Please sign in to leave a comment.