processFields image correction

  • 66 Views
  • Last Post 10 May 2017
  • Topic Is Solved
Sven Perkmann posted this 27 April 2017

Is there any way to setup image correction before processFields is being processed? The field coordinates provided by us are based on a template, but images might be provided in a slightly different angle or distance.

Is there a way to have the image cropped and de-skewed automatically and consistently so processFields hits the right target?

Order By: Standard | Newest | Votes
Oksana Serdyuk posted this 10 May 2017

I apologize for the delay with response caused by the state holidays in Russia.

When you use the processFields method there is no possibility to adjust image preprocessing settings. By default in this scenario the image files are subjected to additional processing, such as:

  • GeometryCorrectionMode = GCM_Auto - image geometry correction will be performed, if it is necessary. ABBYY will determine automatically, whether the processed document is a photo and if it is, will perform geometry correction.
  • ResolutionCorrectionMode = RCM_Auto - which means that if resolution of the image is suspicious, ABBYY should automatically detect image resolution and change it.
  • CorrectSkew = TSPV_Auto - ABBYY will automatically determine if this processing mode should be used, depending on the situation (image characteristics, etc.).
  • CorrectSkewMode = SkewByHorizontalText + SkewByVerticalText - the image skew angle is corrected based on horizontal and vertical text lines.

Unfortunately, Cloud OCR SDK does not support automatic layout analysis for field-level recognition, so you need to set up some certain recognition parameters for the fields to "tell" the service that the text to be recognized has the certain format. The coordinates of the fields also must be specified manually. This approach is applicable, only if the recognized documents are well-structured, have a known layout and are not different from each and other.

Sven Perkmann posted this 10 May 2017

Gotcha, thanks for the detailed response!

Close