skvark / textractor Goto Github PK

View Code? Open in Web Editor NEW

20.0 6.0 6.0 244 KB

OCR application for Sailfish OS. Based on Tesseract OCR engine and Leptonica image processing library.

License: MIT License

QML 58.23% QMake 1.96% C++ 39.81%

tesseract-ocr qml leptonica c-plus-plus sailfishos

textractor's Introduction

Textractor

Textractor is an OCR application for Sailfish OS. Main features:

OCR can be run on:

an image taken with the app
an image selected from the device
a PDF file (one or multiple pages)

Cropping is supported in any reasonable quadrilateral arrangement and perspective correction is applied for the selection. User has access to advanced image preprocessing settings.

Found text can be edited or copied to clipboard. As SFOS is a true multitasking OS, the whole OCR process can be run on background while user can use the device for other purposes at the same time.

Documentation and Help

Textractor Documentation

Environment and building

To be able to build this, follow this Gist to setup the environment correctly: https://gist.github.com/skvark/49a2f1904192b6db311a

In short:

Add my repositories containing Tesseract OCR and Leptonica to the build machine targets.

Preprocessing

Tesseract OCR is just plain engine so Leptonica is used for preprocessing the image.

Currently following steps will be done before the image is passed to the engine for recognition:

Image is first opened using QImage, dpi is set to 300, image is rotated according to device angle and the image is saved in jpg format.
Load the jpg image with Leptonica and convert the 32 bpp image to gray 8 bpp image
Unsharp mask
Local background normalization with Otsu's algorithm
Skew angle detection and rotation (Leptonica decides if the image needs to be rotated)

After those steps the image is passed to the Tesseract.

Test image and result

Original:

Preprocessed

Extracted text:

This is a lot of 12 point text to test the
ocr code and see if it works on all types
of file format.

The quick brown dog jumped over the
lazy fox. The quick brown dog jumped
over the lazy fox. The quick brown dog
jumped over the lazy fox. The quick
brown dog jumped over the lazy fox.






 D R I N K  COFFEE
L Do Stupid Faster
 With More Energy

textractor's People

Contributors

Stargazers

Watchers

Forkers

quent-in wyyrepo sailfish-os-apps ibalram smatkovi

textractor's Issues

Rotate images which are taken with Jolla's camera app

If user tries to process an image which was taken with Jolla's camera app and the orientation was not standard, image will appear in wrong orientation after preprocessing since afaik Jolla's camera app only adds exif orientation information and does not actually rotate the image and seems that Leptonica does not support exif data.

Fix: find a way to read the exif data and rotate the image before preprocessing.

Translation

Hi,

I really like the potential of this app, and would like it even more if I could have it in my native language. I took a look at the .ts file and found that it is far from complete. Are you planning to update translation file and implement localisation? If so, I would be the first to translate.
Great job anyway. Thanks!

Check that common image file formats work

Currently only jpeg is confirmed to be working. This may need changes to Leptonica spec file to be able to build it with additional image formats.

Unload camera after taking a picture

Currently the camera is not unloaded and this makes the app consume too much power if user pushes the app to the background after recognition.

Skip image processing

Is it possible to completely skip the image processing? Send the image directly to tesseract?

Thank you for your work,
Cosmin Popescu.

Write documentation / user guide at least for settings page

Settings page contains complex parameters. Write guide how to adjust them.

Additionally split the settings page to advanced and basic settings.

Add ability to extract text from scanned pdfs

Scanners often create pdf-documents when you use them for copying. It would be real handy if the text in those pdfs would be available.

Add optional ability to translate extracted and possibly edited text via some online translation tool.

Google code shutdown, data files downloading needs refactoring

Due to Google code shutdown Tesseract OCR codebase along other data was moved to GitHub. The data will be archived to Google Code archive, but it will be more future proof to refactor the language data downloading so that it will fetch the data from GitHub according to the release of Tesseract.

3.04 contains already a set of new language data files and Textractor still uses the 3.02 version files.

Todo:

figure out the easiest way to download the .traineddata files from GitHub: https://github.com/tesseract-ocr/tessdata
refactor codebase to support this new method

Error after cropping on SFOS 2.0

Leptonica throws error:

findFileFormatStream: failed to read first 12 bytes of file

Something has changed in new update. Cropping itself seems to be working...

Some icon files missing on SFOS 2.0

Following icons are missing:

icon-camera-automatic
graphics-cover-camera

Fix: try to find icon which exists on SFOS 2.0 and in earlier systems too.

Save preprocessing result or overwrite old file?

Currently Textractor overwrites the old preprocessed image when new recognition is run. Do we need to save all the results or is it enough to see the latest preprocessed image?

Add orientation locking functionality to camera page

If the phone is laying flat (camera pointing down) it is pretty much impossible to detect (or to know/guess) the correct orientation for the image. Solution is to add button to camera page which makes it able to toggle manual lock on different orientations.

Stuff needed: icon for all 4 orientations.