ideally, all the operations get performed on the raw data, then the information will be extracted; To ensure best possible recognition results:
- remove alpha channel
- white balance
- rotate
- deskew
- crop to uniform whitespace
- thresholding (otsu) - binarization
- testCase: 4 corners are white?
- noise removal
- gaussian blur slightly
- remove initial comments in libraries
- upgrade to opencv4.0.1
- switch to nieblack-alg implementations
class ImgObject has: