I am trying to estimate the real world distance (in metres) between two points in a perspective image using an uncalibrated camera. However, the dimensions of an object in the image are known.
I thought of using pixel per metre, but, being a perspective image, that did not seem like a viable approach. I believe I need the camera matrix from the image (maybe using the known object) and compute the 3D coordinates of the point, then simply compute the Euclidean distance between the points.
If that is the right approach, how may I do it? If there are any alternatives to this approach, kindly let me know.