Coordiantes of textblocks on a tilted image taken by mobile

  • Last Post 07 July 2015
  • Topic Is Solved
Sarath Babu posted this 06 July 2015

What will be the output xml produced when I perform processImage on a tilted image. Will the text blocks coordinates are also in tilted form ? Or are a going to return the text blocks as horizontally aligned.

My doubt is, if we use a image that is almost correctly aligned, as a base image. and when ever another tilted image is converted to xml we could use this xml by changing the coordinates values of the xml result, so that the result will be same as that of the result which will be given for a corrected image(image correctly aligned as base image).

the advantage is we don't need to do any preprocessing on tilted image to make align with the base image.

Sorry I could not upload any images. I haven't started anything I just trying to figure out solutions for doubts in my mind.

Order By: Standard | Newest | Votes
Oksana Serdyuk posted this 06 July 2015

ABBYY Cloud OCR SDK supports automated image de-skewing. Also there are two available XML export formats for the processImage method:

  • xml – this format allows you to get text with its original image coordinates,
  • xmlForCorrectedImage - this format allows you to get text with coordinates related to the corrected image (deskewed, rotated, etc.), not the original.

So, if I understand your question correctly, you should always use the xmlForCorrectedImage format for extracting coordinates both for a base image and tilted images.

Sarath Babu posted this 06 July 2015

So if I use xmlForCorrectedImage can I assume that the difference between the result for base image and tilted image will be just scaling factor.

Oksana Serdyuk posted this 07 July 2015

Yes, it should be so.