0

Is there any problem with passing image points to OpenCV's solvePnP function which do not lie inside the image itself? This includes points with negative coordinates.

If have read some tutorials/ answers where it was mentioned implicitely that the point should lie inside the frame but i can't think of any reason why this should be required.

  • how do you know the out-of-image points? I think they have to be part of the same pinhole camera model (same calibration etc.) then in theory it could be ok. But not sure about the acutual algorithm. – Micka Oct 13 '19 at 14:57
  • i do some geometric calculations, such as fitting lines through detected points and calculating intersection points – Commodore Yournero Oct 13 '19 at 20:58

0 Answers0