Skew detection and correction in scanned images.
Project description
hough - Skew detection in scanned images
Hough finds skew angles in scanned document pages, using the Hough transform.
It is oriented to batch processing, and can make use of multiple cores or an optional CUDA backend. (It can be very compute intensive!)
Installation and usage
Installation
pipx install hough
Or, if you have a supported GPU and have installed CUDA (currently 12.x supported):
pipx install "hough[cuda]"
If you don't use pipx, other methods such as pip should work fine, just create a virtual environment first.
Usage
To get started right away, here's some examples.
Generate angles (in CSV form) for a bunch of TIFF images:
hough analyse in/*.tif
The same, but for a PDF file, and display a histogram at the end:
hough analyse --histogram Able_Attach_Sep83.pdf
The same, but show more information while running:
hough --verbose --histogram Able_Attach_Sep83.pdf
The deskewing results are placed in a results.csv file created under the out/<timestamp> directory, which is created at invocation time. Here's an example:
"Input File","Page Number","Computed angle","Variance of computed angles","Image width (px)","Image height (px)"
"/home/toby/my-pages/orig/a--0000.pgm.tif",,-0.07699791151672428,0.001073874144832815,5014,6659
"/home/toby/my-pages/orig/a--0001.pgm.tif",,,,5018,6630
"/home/toby/my-pages/orig/a--0002.pgm.tif",,0.24936351676615068,0.005137031681286154,5021,6629
"/home/toby/my-pages/orig/a--0003.pgm.tif",,,,5020,6608
"/home/toby/my-pages/orig/a--0004.pgm.tif",,-0.037485115754500545,0.025945115897015238,5021,6616
The program should work on various image input formats, and with both grey scale and RGB images. Hough works best with images ≥300dpi.
Here's a histogram sample:
=== Skew statistics ===
0.00° - 0.10° [57] ████████████████████████████████████████
0.10° - 0.20° [39] ███████████████████████████▍
0.20° - 0.30° [30] █████████████████████
0.30° - 0.40° [30] █████████████████████
0.40° - 0.50° [11] ███████▊
0.50° - 0.60° [11] ███████▊
0.60° - 0.70° [ 3] ██▏
0.70° - 0.80° [ 4] ██▊
0.80° - 0.90° [ 0]
0.90° - 1.00° [ 1] ▊
1.00° - 1.10° [ 1] ▊
1.10° - 1.20° [ 0]
1.20° - 1.30° [ 1] ▊
1.30° - 1.40° [ 1] ▊
1.40° - 1.50° [ 1] ▊
1.50° - 1.60° [ 2] █▍
1.60° - 1.70° [ 0]
1.70° - 1.80° [ 1] ▊
1.80° - 1.90° [ 2] █▍
1.90° - 2.00° [ 0]
Samples: 195
50th percentile: 0.20°
90th percentile: 0.55°
99th percentile: 1.77°
Command line options
You can list them by running hough --help:
Usage: hough COMMAND
╭─ Commands ─────────────────────────────────────────────────────────────────────────────╮
│ analyse Analyse one or more files for deskewing. │
│ histogram Show a histogram of rotation angles from a previous analysis. │
│ process Fully analyse and rotate one or more files. │
│ rotate Rotate one or more files that have previously been analysed. │
│ --help -h Display this message and exit. │
│ --version Display application version. │
╰────────────────────────────────────────────────────────────────────────────────────────╯
Or ask for help for a specific command, e.g. hough analyse --help:
Usage: hough analyse [ARGS] [OPTIONS]
Analyse one or more files for deskewing.
╭─ Parameters ───────────────────────────────────────────────────────────────────────────╮
│ * FILES --files One or more files to analyse for deskewing. [required] │
│ DEBUG --debug Save intermediate results in debug/ under out folder. │
│ [default: False] │
│ VERBOSE --verbose Print status messages instead of progress bar. [default: │
│ False] │
│ OUT --out Use the specified path for results and post-rotated files. │
│ [default: out/TIMESTAMP] │
│ WORKERS --workers Number of workers to run simultaneously. [default: 4] │
│ HISTOGRAM --histogram Display result summary as histogram after processing. │
│ [default: False] │
╰────────────────────────────────────────────────────────────────────────────────────────╯
Examples
Just about all of these files have been deskewed this way.
Getting the best results
NOTE: This is a beta product!
There's a few guidelines you should follow to get the best deskewing results from your document scans:
- Bilevel (black-and-white) bitmaps will produce lower quality results. For best results, scan to greyscale or RGB first, deskew with Hough, then reduce the colour depth to bilevel if desired.
- Hough deskewing is an inexact process, with many heuristics discovered by trial and error. Hough may not work well on your material without tuning and further modification. (We'd love your pull requests!)
Debugging output
You can spy on Hough's attempts to perform deskewing by passing the --debug
flag on the command line. The generated images, and any detected lines in them,
are placed in the out/<timestamp>/debug/ directory.
Note that Hough cannot always determine a skew for a page (e.g. blank pages in particular), and will very occasionally get the skew wrong (depending on source material). It's worth reviewing these images if Hough makes a bad decision on your scans. Please submit these files along with the original image when filing an issue!
Recommended scanners
The authors have tested this software with output from the following scanners:
- Fujitsu fi-4530C, USB
- Fast
- Cheap on eBay
- Requires a Windows XP VirtualBox for drivers
- Brother ADS-2700W, USB + Ethernet + WiFi
- Fast
- Can scan directly to the network or to a memory stick
- Factory reconditioned models stilll available (March 2020)
- Very low skew out of the box
- Epson WF-7610, USB + Ethernet + WiFi
- 11"x17" and duplex capable
- Can scan directly to the network or to a memory stick
Developing
First, clone this repo.
You'll need to install Poetry, then run:
poetry sync --with dev # or --with dev,cuda if you have CUDA installed
poetry self add 'poethepoet[poetry_plugin]' poetry-plugin-shell
Do some work, then run the pre-commit checks and tests with:
poetry run pre-commit
poetry poe test
License notice
This file is part of "hough", which detects skew angles in scanned images
Copyright (C) 2016-2020 Toby Thain <toby@telegraphics.com.au>
Copyright (C) 2020-2025 Joan Touzet <wohali@apache.org>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file hough-0.3.1.tar.gz.
File metadata
- Download URL: hough-0.3.1.tar.gz
- Upload date:
- Size: 29.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.12.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
2d785ff3e4669401569d7fb9538c00e3198029fe5383e1e0c369266a6b55b2d6
|
|
| MD5 |
f23c124c84c40757e5ff5fbaef7a743f
|
|
| BLAKE2b-256 |
f77e1526237176c2f8086253892239a723c8cfef621c39b72b0cb2b1e578ba09
|
Provenance
The following attestation bundles were made for hough-0.3.1.tar.gz:
Publisher:
publish.yml on wohali/hough
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
hough-0.3.1.tar.gz -
Subject digest:
2d785ff3e4669401569d7fb9538c00e3198029fe5383e1e0c369266a6b55b2d6 - Sigstore transparency entry: 208857983
- Sigstore integration time:
-
Permalink:
wohali/hough@8ba0ce9ead0d2bb283186ea9d97ae7588408ab6c -
Branch / Tag:
refs/tags/untagged-390eb43e96614133372d - Owner: https://github.com/wohali
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@8ba0ce9ead0d2bb283186ea9d97ae7588408ab6c -
Trigger Event:
release
-
Statement type:
File details
Details for the file hough-0.3.1-py3-none-any.whl.
File metadata
- Download URL: hough-0.3.1-py3-none-any.whl
- Upload date:
- Size: 27.1 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.12.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
3b937b96ad0eab507e0a49b41a4f8d0cfea6c9d0b935c9cdd7e308aa52640064
|
|
| MD5 |
82b3e11d8bc374fac9a6717a944f2f28
|
|
| BLAKE2b-256 |
b2692dd289cfa611bbeba3413088c568f3ae6c786e4312fffc3db4379ddc2210
|
Provenance
The following attestation bundles were made for hough-0.3.1-py3-none-any.whl:
Publisher:
publish.yml on wohali/hough
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
hough-0.3.1-py3-none-any.whl -
Subject digest:
3b937b96ad0eab507e0a49b41a4f8d0cfea6c9d0b935c9cdd7e308aa52640064 - Sigstore transparency entry: 208857985
- Sigstore integration time:
-
Permalink:
wohali/hough@8ba0ce9ead0d2bb283186ea9d97ae7588408ab6c -
Branch / Tag:
refs/tags/untagged-390eb43e96614133372d - Owner: https://github.com/wohali
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@8ba0ce9ead0d2bb283186ea9d97ae7588408ab6c -
Trigger Event:
release
-
Statement type: