Computer vision: Write a document scanner in Python 3

Completed Posted 7 years ago Paid on delivery
Completed Paid on delivery

The job is to write a Python 3 module that implements the following algorithm:

1. Read a .png or .jpeg image of a document or letter.

2. Find contours of the letter, similar to what is shown in this article:

[login to view URL]

However, instead of assuming that the letter is rectangular, this step should be more sophisticated: It should find better contours, like this:

[login to view URL]

3. Unskew the letter into a rectangular hull. It is NOT ok to just add white space; instead, the whole document should be "undistorted" by this step. For example, the original image may look like this, and the distortion-correction should still work:

[login to view URL]

4. If needed, correct the text angle like this:

[login to view URL]

5. Apply OCR / create a searchable PDF.

Other requirements:

- The original colors should be retained in the result.

- The Python 3 API should be similar to the following:

import paperscan as ps

with open('[login to view URL]') as fp:

img = [login to view URL](fp)

hull = [login to view URL](img)

page = [login to view URL](img, hull)

angle = ps.text_angle(img, page)

page = ps.fix_angle(img, angle)

text = [login to view URL](page)

with open('[login to view URL]', 'w') as fp:

ps.write_pdf(fp, page)

Computer Graphics Python Software Development

Project ID: #11555619

About the project

2 proposals Remote project Active 7 years ago

Awarded to:

iitmshanker

Hello, I have worked on several computer vision problems like salient part detection in images, face recognition, text recognizer etc Please ping if you are interested in working with me. thanks,

€200 EUR in 5 days
(23 Reviews)
5.2