Image preparation - Githubissues

jimmykarily / open-ocr-reader

An application to allow blind or visually impaired people to read books (the paper ones).

0 stars 0 forks source link

Image preparation #2

Open jimmykarily opened 2 years ago

jimmykarily commented 2 years ago

Input: an image of a book (or book page) Output: a image ready to pass text block detection or OCR

There are filters that can be applied to a photograph of text that can make it more suitable for OCR.

Some of them:

deskew (align the document vertically)
improve contrast
remove noise (spots etc)
black and white
more: https://gitlab.gnome.org/World/OpenPaperwork/libpillowfight#available-algorithms
https://github.com/avelino/awesome-go#images
4 point image transformation: https://gouravkumar.dev/articles/golang-four-point-image-transformation-gocv-opencv/
Detect contours: https://pyimagesearch.com/2014/04/21/building-pokedex-python-finding-game-boy-screen-step-4-6/
Detect objects with yolo3: https://github.com/wimspaargaren/yolov3
https://becominghuman.ai/how-to-automatically-deskew-straighten-a-text-image-using-opencv-a0c30aed83df (code here: https://github.com/JPLeoRX/opencv-text-deskew/tree/master/python-service)