Principal points of rectified images in opencv

0
votes

In Open, I have calibrated my cameras using stereo calibration function. then I have rectification left and right images. Until now, everything is working well.

But I don't know how to get the principal points of images for the left and right camera. I need them for triangulation the features from left and right images.

I have extrinsic and intrinsic parameters of the camera but I don't know how to calculate these principal points of rectified image.

1
opencvtriangulationstereo-3dcalibrationprincipal
When you do rectification you get projection matrix in the new (rectified) coordinate systems, P(0, 2) and P(1, 2) should correspond to cx and cy(principal points) . Proof: P = [KR, -Kt], since rectified therefore R = I, hence P = [K, -Kt] – Optimus 1072
Thank you! It is exactly would you have mentioned! I have once more question if someone can help me would be great. Why we have fx and fy separately? in which condition they will be same fx=fy? for my triangulation I need only f which is focal length. – A. Fasih
so pixels are not square , they are generally rectangle and hence we have fx and fy. Generally the difference between fx and fy is not too much, as far as I remember you can use any (or may be their mean), it wouldn't matter much. – Optimus 1072
If your both questions are answered then should I explain it as an answer so that you can accept it and we can close this question? – Optimus 1072
Thank you very much. Your answer was very useful for me. – A. Fasih

1 Answers

0
votes

When you do rectification you get projection matrix in the new (rectified) coordinate systems, P(0, 2) and P(1, 2) should correspond to cx and cy(principal points) . Proof: P = [KR, -Kt], since rectified therefore R = I , hence P = [K, -Kt].