skew angle and transformation matrix

  • Last Post 2 weeks ago
tarun73 posted this 3 weeks ago

Hi ,

1. The image returned through the "pdfsearchable" option is very degraded image. is there an option to get a good quality image. 

2. Is there any way to know the 4 corners (coordinates) or edges which i can use to remove the skew from the original image myself. 

3. Is there any difference in the output of text recognition, in both cases as below

   a) sending small blocks of text from an image and later concatenating the text .

   b) Send the whole image.


Nikolay Krivchanskiy posted this 2 weeks ago


1. Could you please specify what you mean by “degraded image”? The best would be if you could also attach the example file or send it to us via email. That would help us to look for parameters to improve recognition.

2. You can export the image into the .xml, using “xmlForCorrectedImage” export parameter. Result will contain the coordinates as it is presented on the corrected image. Also you can export into .xml with coordinates, corresponding to the original image, using “xml” export parameter instead. Coordinates of page will be stored in “page” tag in .xml output. For more information about .xml export scheme, please refer to the following link to our website.

3. There might be a slight difference, but generally technology already splits image into blocks to further recognize it block-by-block. Please note that recognition quality is highly dependent on the quality and resolution of an image, but also can be influenced by moderate text size/brightness/sharpness difference on the same image.