A programmatic approach to finding document edges in images

It is very common that a valid ID is required when signing up for websites, services and apps. As more and more of our daily tasks move to smartphones, users should have the ability to simply take a picture of their form of identification, regardless of the background and orientation. The proposed algorithm allows to process images coming from cameras, identifying the borders of paper documents, to then crop and straighten the selected area to fit a rectangular container. The proposed solution would not use artificial intelligence models, as it would be extremely difficult to gather a large enough dataset of valid IDs, even after considering data augmentation techniques.