QRFx - finding the best way to waste my time

not a waste of time though, learnt a lot but... a project but only customized for my use case.

this project is not made to fix any broken qr image.
i haven't taken generalization of the algo into consideration. I solely made this project to fix a qr code (single) I already had which wasn't working.

original QR code, doesn't work.

i had this qr code lying around since a few years... yes i had a plan to recover it since then, i was given an opportunity to do any project and implement anything for a thing i applied, so i decided to proceed with this, and i surprisingly picked up opencv quite well in just 5 days and got the qr code working.

technical specifications

what i tried

i tried gaussian blur, simple blur using kernels, filters, bilateral filtering, binarization, canny edge detection, gap filling, contours, flood fill, otsu threshold too.

this one above works, and is scannable.

actual current pipeline (for locating it manually)

these all steps are written for me to try to get an algorithm made for detecting a qr code's location in an image.

• input (source.png) image gets read in grayscale
• gaussian blur is applied 5 times
• the output gets binarized with 170 threshold
• then canny edge detection is applied on the binarized image
• then a custom algo spots right angles to try and detect a qr code in the inputted image. (and cropping it)
  we can crop a qr code if we just get 2 out of 3 finder patterns detected.

future additions can be then digitizing the qr code with contour polygon approximations overlayed and converting it to the nearest polygons possible and getting a decent qr code. just for the fun, I don't think it has any practical uses.

ALSO, please don't judge my coding skills with this (they are way worse /s), i write trash code but ay atleast it works, i skipped on the documentation part because no one else is gonna use this abomination of a code apart from me.