Baudot-Murray/CCIR 476 decoder

This repository demonstrates a complete workflow for digitally decoding Baudot-Murray/CCIR 476 encoded teleprinter tape from microfilm images using computer vision techniques. The Baudot-Murray code (also known as CCIR 476) was a 7-bit character encoding system used in early teleprinter systems.

How it works

The main processing pipeline is implemented in code_extraction.ipynb and consists of the following steps:

1. Image Preprocessing

• Loads the microfilm image from 'Data/Baudot-Murray-example_cropped.jpg'
• Converts to grayscale and applies binary thresholding
• Uses probabilistic Hough line detection to identify the tape structure
• Creates a convex hull around detected lines to isolate the code area
• Crops the image to focus on the encoded data

2. Grid-Based Character Recognition

The system recognizes that Baudot-Murray tape uses a specific grid structure:

• 26 rows of encoded characters
• 16 columns total, with only 14 used (2 groups of 7, separated by 2 spacing columns)
• Each cell represents a binary bit (hole = 1, no hole = 0)

3. Binary Extraction

For each grid cell:

• Samples the pixel values in that region
• Converts to binary based on average brightness (threshold at 177/255)
• Builds a binary representation of each character row

4. Character Decoding

• Splits the data into two 7-bit sections (left and right groups)
• Flips the bit order to match standard Baudot-Murray encoding
• Uses the lookup table in 'Data/binary_to_ascii.json' to convert binary patterns to readable characters

Key Features

• Automated tape detection: Uses Hough line detection to automatically locate and crop the tape area
• Grid-based sampling: Precisely samples each bit position using calculated grid coordinates
• Visual feedback: Displays the processing steps including detected lines, grid overlay, and extracted binary values
• Complete character set: Supports the full Baudot-Murray character set including letters, numbers, and control characters

Usage

Install the dependencies:

pip install -r requirements.txt

Run the code_extraction.ipynb notebook cell by cell to see the complete processing pipeline. The notebook will:

1. Load and preprocess the microfilm image
2. Detect and extract the tape area
3. Apply grid-based character recognition
4. Display the decoded text output

Sample Data

The repository includes:

• Data/Baudot-Murray-example_cropped.jpg - A cropped sample of teleprinter tape from microfilm
• Data/binary_to_ascii.json - Complete Baudot-Murray character encoding table
• Data/output.png - Example output visualization

This workflow can be adapted for other historical teleprinter tape formats and provides a foundation for digitizing archived communication records.

License

The project is licensed under the MIT License, allowing free use, modification, and distribution.