Paper Deshredder

Reconstructing shredded documents using computer-vision

How it works

Paper Deshredder is a python prorgram that takes a picture of paper shreds as input and gives you a reconstructed document as an image. Paper Deshredder uses computer vision and image processing to identify the shreds in the image, re-orient them, and stitch them back together in to correct order.

Steps

1. Run a saturation convolution on the input image do detect shred boundaries
2. Identify each shred using the convolution image, extract the shreds, and re-orient the shreds
3. Analyze each shred and stitch the reconstructed image

Constraints

• The input must be a single image containing all shreds to be considered
• The shreds must be placed on a saturated surface (distinctly not black or white)
• The shreds must have relatively straight edges and be rectangular in shape
• The shreds should be "fairly upright", such that the angle between the bottom of the shred and the bottom of the picture is less than 45 degrees
• The content of documents to deshred must be grayscale in color

Freedoms

• Shreds can be placed in any orientation except for previously defined constraints

Running

• To execute this program, run main.py with the --help option to list arguments and options.

NOTE: YOU MUST BE IN THE ROOT DIRECTORY WHEN YOU RUN python src/main.py