Revision history [back]

Hi, everyone.

I have one question about FocalLenght in documentation in CameraCalibration. No we have formula like this:

Why do we multiply by Fx and Fy in the last two rows, but not in the second and third rows? In my opinion we should use this rows:

......
x' = Fx * x / z
y' = Fy * y / z
......
u = x'' + Cx
v = y'' + Cy

Am I right? Or I do not understand something?
Can somebody help me to find error?

Thank you, Maxim

Hi, everyone.

I have one question about FocalLenght in documentation in CameraCalibration. No Now we have formula like this:

Why do we multiply by Fx and Fy in the last two rows, but not in the second and third rows? In my opinion we should use this rows:

......
x' = Fx * x / z
y' = Fy * y / z
......
u = x'' + Cx
v = y'' + Cy

Am I right? Or I do not understand something?
Can somebody help me to find error?

Thank you, Maxim

Hi, everyone.

I have one question about FocalLenght in documentation in CameraCalibration. Now we have formula like this:

Why do we multiply by Fx and Fy in the last two rows, but not in the second and third rows? In my opinion we should use this rows:

......
x' = Fx * x / z
y' = Fy * y / z
......
u = x'' + Cx
v = y'' + Cy

Am I right? Or I do not understand something?
Can somebody help me to find error?

Thank you, Maxim

Hi, everyone.

I have one question about FocalLenght in documentation in CameraCalibration. Now we have formula like this:

Why do we multiply by Fx and Fy in the last two rows, but not in the second and third rows? In my opinion we should use this rows:

......
x' = Fx * x / z
y' = Fy * y / z
......
u = x'' + Cx
v = y'' + Cy

Am I right? Or I do not understand something?
Can somebody help me to find error?

Thank you, MaximEDIT:

For simplicity lets consider only Z and X. Lets Y = 0 everywhere.

D(x) — it is distortion polynom
x — original point coordinates (X and Zx axises)
x' — it is undistorted (ideal) coordinates (X and Zx axises)
x'' — it is distored (real) coordinates (X and Zx axises)
v — it is distored (real) coordinates (V and Zv axises)

In my opinion to calculate image coordinates (V and Zv axises) using camera coordinates (X and Zx axises) we should use these equations:
x' = f * x / z
x'' = D(x')
v = x'' + c

There is my mistake?
Why in OpenCV documentation we multiply by focus at the end, but not at the begin?