hocr-resizer

Command-line utility for resizing the coordinates in an hOCR (html-formatted ocr text) file.

Use Case:

• lets say that we've already done the following:
  • tesseract was used to generate hOCR data from a set of high-resolution images
  • hocr-pdf was used to generate a PDF document that contains:
    • one visible image per page
    • ocr text in an invisible layer that makes the images searchable
• what if:
  • the size of the PDF is too big
• we can:
  • use ImageMagick to resize all images to a lower resolution
• the problem:
  • the coordinates in the hOCR files no-longer correspond to the dimensions of the images
    • the aspect ratio hasn't changed
    • the width and height of the images (in pixels) have decreased

Existing Solution:

• this problem has been discussed in an issue in the hocr-tools repo
• as mentioned in an issue comment
  • hocr_resizer.rb is a Ruby script that can rescale the coordinates in a hOCR file
  • hocr_resize.rake is a Ruby rakefile to call the script

Reason for Yet-Another Solution:

• I fkn hate Ruby, and its enormous non-portable runtime
• ..why not?

Installation:

npm install --global @warren-bank/node-hocr-resizer

Usage:

hocr-resizer <options>

options:
========
"--help"
    Print a help message describing all command-line options.

"-v"
"--version"
    Display the version.

"-w" <integer>
"--width" <integer>
    [required] Width of new/resized image.

"-h" <integer>
"--height" <integer>
    [optional] Height of new/resized image.
    Default: calculated from old aspect ratio and new/resized width.

"-i" <filepath>
"--input" <filepath>
    [required] Filepath to input hOCR file.

"-o" <filepath>
"--output" <filepath>
    [optional] Filepath to output hOCR file.
    Default: overwrite input hOCR file.

Example:

• overwrite an hOCR with updated coordinates based on the same aspect ratio and a new image width of 1275px (ie: 150dpi @ 8.5"):

    hocr-resizer -w 1275 -i '/path/to/file.hocr'

Legal:

• copyright: Warren Bank
• license: GPL-2.0