I think the change is legit. However, it brings a slightly bigger issue. We also have a few other places, where the principal point is treated the similar way. Namely, getOptimalNewCameraMatrix and stereoRectifyUncalibrated (maybe, somewhere else, as well). So, these functions should, probably, be modified accordingly.

Also, we'll need to add a couple of tests (the provided sample can simply be refactored into a test).

@vicproon, do you care to take a look at this?

@vicproon, @snosov1, wow, frankly speaking, I did not expect such a dramatical change because of extra "-1" added or removed. It still looks like a magic to me. Still, I'm fine to move from nx-1, ny-1 to nx and ny. 👍

@snosov1 I am not sure i can make in immediately, but i sure can take a look at this a bit later. How about creating an issue for these updates?

@vpisarev this happens because when you calculate camera matrix for 4 times smaller picture, this extra "-1" becomes "-4".