Image Resolution: What are the optimal settings for OCR

What is the best image resolution when using ABBYY OCR technologies?

  • When scanning regular texts with font size 8-10 points, it is recommended to use 300 dpi resolution for OCR. All ABBYY technologies are tuned for that resolution.
  • For smaller font text sizes (8 points or smaller) we recommend using A 400-600 dpi resolution.
  • If scans have a smaller resolution, for example, 200 dpi, then 10 point font size will occur to be too small for a good quality OCR result. To compensate for the “missing” pixels, the image will be internally scaled up (up to 400 dpi). The low image quality caused by the low resolution may lead to lower OCR accuracy. It can as well cause speed degradation as uncertainty in character picture produces more recognition variants (which take more time to process).

Recommendations and limitations of OCR technology for a character size in pixels

  • 1-byte (simple script) languages like English, Russian
    • Recommended: 20
    • Minimal: 12
  • 2-byte (complex script) languages like Japanese, Chinese
    • Recommended: 25
    • Minimal: 22
    • Maximum: 60 (for body text, for heading there is no limit)
  • 1-byte (complex script) languages like Thai, Hebrew, Arabic
    • Recommended: 20
    • Minimal: 12

Page layout analysis requires a small character size in a text string bigger than 1 millimeter.

Recommendation for color, grey or black & white images

  • ABBYY technologies use color information for detecting areas and objects on the image.
  • If complex layouts need to be processed, it is recommended to use color or at least, greyscale images.
The optical character recognition is always executed on a bi-tonal image, that only contains black & white. To achieve optimal OCR results, it is, therefore, crucial to generate a good, suitable binary image. In order to receive the highest recognition quality, ABBYY technology is not just using “simple” binarization, but own adaptive binarization technology.

What is the largest supported font size for OCR?

  • The largest font ABBYY OCR can handle is 5cm or 140pt.

What is the largest image size in pixels?

  • ABBYY products can open image formats up to 32512*32512 pixels.
  • Bigger images have to be cut and the segments have to be processed separately

What is the maximum image file size?

  • ABBYY Technology v.8.x and older were able to open image files up to 2 GB.
  • Starting from V9.0 this limit doesn't exist anymore. 

Character-Shapes with different Image Resolutions

  • The lower the image resolution has a strong influence on the way how a single character is “built up”.

  • The image resolution has a real impact on the OCR quality that can be archived - below a sample of small text of a fax, the problems you can see here
    • Text is not on a straight line
    • Characters are “squeezed” and glued together
    • The resolution is by far too low for the classifiers
    • → you (as a human) might be able to read it when you know the language and the context.

  • But when zooming in further, on the pixel level, you probably will fail - so do the algorithms :-(

Additional information

Was this article helpful?

3 out of 4 found this helpful

Have more questions? Submit a request

Comments

0 comments

Please sign in to leave a comment.

Recently viewed