Looking for experienced C++ Computer Vision programmer to develop text detection service.
Input data is jpg/png image of text file.
Considering file could be low-quality one (cell picture of original document, for example)
it means that :
1) it could be rotated in 2D (around)
2) turned in 3D ( one edge (or corner) of image is farther then another (not more then 20 degree angle usually))
3) One part of image could be darker then another
As we understand, this means you need to apply some type of transformations if needed, such as:
- turning it
- affine transformation
- some sort of cleanup
Next step is text detection. We need to have borders coordinates of each word in present text (in context of rational possibility based on quality of input image, of course )
As work result we wish to have source code of program.
To check work quality we wish to have compiled version of program ( or internet service maybe ) , to test it with some our images.
Our input file is InputPic.jpg in attachments
In a perfect world, after all the processing we want to obtain something like in SamplePicMod.jpg file ( TWO STRINGS AT THE TOP OF PIC WITH EACH WORD DETECTED )
And to get the coordinates of the words borders (each corner) related to top left corner of the pic.
We will start working with you after you show at least affine transformations of the pic, noize removal and make it "face-forward". As we understand, for this kind of work you should use affine based on 3 points. As i imagine, you can pick up this points based on strings lines position (maybe not)
Next, we extremely wish to get very well-documented source code. Thank you.
I am looking for a mix of experience and value