Book Scanner Software

Introduction

This page is for document efforts beginning in 2017 to create a software OCR pipeline.

This is a continuation of efforts by the Digital Archivists Working Group to make the book scanner more convenient to work with.

---

Platform

The current hardware platform consists of the following elements-

• A wooden book scanner\ frame, built by members of the Digital Archivists working group
• Two cameras (currently Canon EOS Rebel T3), mounted in the book scanner
• Lighting, glass plates, book platen
• USB cables connecting cameras to computer
• Apple Mac Mini (mid 2010: Core2Duo 2.4 GHz, 8GB RAM, NVIDIA GeForce 320M 256MB)
• macOS Sierra 10.12.3

---

Intent

The intent of this effort is provide an OCR facility to help create digital documents from the pages photographed by the book scanner platform. Aspects of this may include-

• Photo manipulation (cropping, rotation, image adjustments)
• Partitioning of photos to provide hints to the software about text versus image regions
• OCR conversion of characters in photos to text files
• Hooking into the OCR API to obtain confidence level or probability data about each image conversion result, possibly for directing the human operator to regions that may need correcting
• Connecting together and automating of any of these aspects

Softwares

Scanner & Camera Control

• CHDK: Camera control package that Steele Nivenson has been hacking on, to the point it's functioning quite well with used, low-cost 16MP mirrorless Canon pocket cameras (PowerShot A2200, A2500), in addition to the EOS DSLR lineup.

Acquisition & Image Post-Processing

• spreads: Project on Github implementing acquisition through OCR, apparently by Johannes Baiter.

Efforts

• 4 March 2017: A session with Tesseract
• 16 March 2017: Tested a Trial Copy of ABBYY FineReader, an image-to-PDF OCR package starting around $120
• 18 March 2017: Install and try out Spreads on a Mac Mini
• 20 March 2017: Continue Spreads Install on a Mac Mini
• 30 May 2017: Test a copy of PDFScanner, a macOS scanner/image-to-PDF package costing about US$16