QR codes were developed for use on assembly lines. To ensure a successful project I would assume that my camera was mounted on an assembly line and I would have these limitations on the scope;
- The system will correctly decode 98% of QR code images.
- The QR codes will all be Version 1 (21 x 21)
- The QR codes will all have the same orientation
- The sides of the QR code will be parallel with the sides of the image
- The QR code will be completely within the image
- The QR code will be as large as possible within the image
- The image will be in Black & White i.e. Grayscale (if possible)
- The image will be at the most suitable resolution (probably the lowest resolution the camera can take)
- The image will be in the simplest format possible (monochrome BMP ?)
- The lighting will be good and the same for all images
The above limits on the project scope result in something achievable that can be done on an Arduino.
Without such limits the project could involve complex rules, maths, geometery, and even AI and neural nets and be well beyond what an Arduino can do.