Scan & OCR scripts

This is a very productive scanning and OCR setup, intended to speed up the scanning process and produce a CBZ file and an archive of extracted text as fast as possible. Just follow these steps:

1. install the required packages
2. plug in your scanner
3. edit config.sh according to your needs (see Configuration)
4. run ./1-scan.sh
5. do any necessary renaming and extra scanning (see Naming convention)
6. run ./2-ocr.sh
7. run ./3-bundle.sh

This setup was inspired by How to scan and OCR like a pro with open source tools. The article also explains a few things not included in these scripts, like how to remove page numbers and unnecessary line feeds. Add these parts in if you need to.

Required packages

In Debian:

sudo apt install sane sane-utils imagemagick unpaper tesseract-ocr

Also, install the Tesseract language package(s) you need. Select from:

• apt search tesseract-ocr-
• https://packages.debian.org/search?keywords=tesseract-ocr-

If you have an old version of Debian, install the newer Tesseract language package(s) from backports. Example (for Debian 9): add deb https://deb.debian.org/debian stretch-backports main to /etc/apt/sources.list, then run:

sudo apt -t stretch-backports install tesseract-ocr-eng

Configuration

Before using the scripts, you must edit config.sh according to your needs. You need to change at least the following options:

1. device: run scanimage -L to find the device id. Ex: device='genesys:libusb:001:004'.
2. width and height: measure the pages' width and height in millimeters. Images will be cropped to this size automatically.
3. first_page and last_page. first_page can be a negative number, if needed (see below).

Other important options are:

4. language: the language setting for OCR must correspond to the document's language. Ex: 'eng' for English, 'ron' for Romanian.
5. rotate: angle for clockwise auto-rotation of every page. Possible values are 0, 90, 180 and 270.
6. resolution in DPI, defaults to 300.

Naming convention

For clarity, we want file names to match page numbers: 001.pnm for page 1, etc. As for the unnumbered pages (covers, inserts, folds, etc), we must name them in a way that preserves page order. This is especially important when generating CBZ files, in which page order is determined by file names. We have two main situations:

1. The cover and first few pages might not be numbered. In this case, set first_page to a negative number. Before reaching page 1, files will be named 000_1.pnm, 000_2.pnm, etc.
2. In case of other unnumbered pages (inserts, folds, etc), skip them on the first run and scan them separately, using the command ./1-scan.sh filename_without_extension. For ordering to be consistent, name the files as in the following examples:
   • If there is an insert between pages 45 and 46, use this convention: 045_0.pnm, 045_1.pnm, 045_2.pnm, 046.pnm. So after the first run, rename 045.pnm to 045_0.pnm, then scan the insert by running ./1-scan.sh 045_1 and ./1-scan.sh 045_2.
   • If leaf 45/46 is folded and actually contains 4 pages, use this convention: 045_1.pnm, 045_2.pnm, 046_1.pnm, 046_2.pnm. So after the first run, rename 045.pnm to 045_1.pnm and 046.pnm to 046_1.pnm, then scan the extra pages in the fold by running ./1-scan.sh 045_2 and ./1-scan.sh 046_2.
   • If leaf -1/0 (the front cover) is folded and actually contains 4 pages, use this convention: 000_1_1.pnm, 000_1_2.pnm, 000_2_1.pnm, 000_2_2.pnm. So after the first run, rename 000_1.pnm to 000_1_1.pnm and 000_2.pnm to 000_2_1.pnm, then scan the extra pages in the fold by running ./1-scan.sh 000_1_2 and ./1-scan.sh 000_2_2.

Important: you must do all renaming before running ./2-ocr.sh!