Hiya, in our app we are trying to give our customers an estimate of the tracking-quality of an image they choose. We don't need the exact algorithm or Wikitude's internal rating system, an approximation is enough.
We know Wikitude is closed source, so we cannot get the quality of a target outside of the target API that takes rather long to communicate with. So we wanted to ask if anyone had any suggestions for algorithms that could be used to determine the quality of a camera image during runtime.
So far we've played around with OpenCV functions like ORB and FAST as well as Good Features to Track. However, the values we acquire from these functions never actually seem to have any relation to the Wikitude target rating (or actual quality).
Has anyone ever worked with anything similar to achieve the same effect? Maybe you could help us out.
Any suggestions appreciated!