Exporting Box Images

[Zack Cohen]

Hi,

I was wondering if anyone knows a way to collect image data from each box.  i.e. where tesseract guesses the character are around each character.  I would like to export each letter as its own image.  Is this possible with the given API?

[Quan Nguyen]

You can GetComponentImages to return the coordinates of the bounding boxes.

https://code.google.com/p/tesseract-ocr/wiki/APIExample

[Zack Cohen]

Thanks!

And then I assume I would use something like magick++ (which I am having a suprisingly hard time getting to work on mac) to crop the images

[Helmut Wollmersdorfer]

Hi,

Am Freitag, 30. Oktober 2015 16:39:17 UTC+1 schrieb Zack Cohen:

Thanks!

You can get it just from the console API at nearly no additional runtime:

$ tesseract page_152.png page_152 -l deu-frak+deu  makebox hocr

This will output three files: page_152.txt, page_152.hocr and page_152.box.

With the data in the box-file you can cut out the areas from the image.

And then I assume I would use something like magick++ (which I am having a suprisingly hard time getting to work on mac) to crop the images

Exactly. And on a mac you should use homebrew to install magick.

 

HTH

Helmut Wollmersdorfer